sign in sign up
<<
Home , Visualization (graphics)

European Research Consortium for Informatics and Mathematics Number 44 January 2001 www.ercim.org

Special Theme: and

Next Issue: April 2001 Next Special Theme: Metacomputing and Grid Technologies CONTENTS

KEYNOTE 36 Physical Deforming Agents for Virtual Neurosurgery by Michele Marini, Ovidio Salvetti, Sergio Di Bona 3 by Elly Plooij-van Gorsel and Ludovico Lutzemberger 37 Visualization of Complex Dynamical Systems JOINT ERCIM ACTIONS in Theoretical Physics

4 Philippe Baptiste Winner of the 2000 Cor Baayen Award by Anatoly Fomenko, Stanislav Klimenko and Igor Nikitin 38 and Visualization of Processes 5 Strategic Workshops Ð Shaping future EU-NSF collaborations in in Moving Granular Bed Gas Cleanup Filter Technologies by Pavel Slavík, František Hrdliãka and Ondfiej Kubelka THE EUROPEAN SCENE 39 Watching Chromosomes during Cell Division by Robert van Liere 5 INRIA is growing at an Unprecedented Pace and is starting a Recruiting Drive on a European Scale 41 The blue-c Project by Markus Gross and Oliver Staadt SPECIAL THEME 42 Augmenting the Common Working Environment by Virtual Objects by Wolfgang Broll 6 Graphics and Visualization: Breaking new Frontiers by Carol O’Sullivan and Roberto Scopigno 43 Levels of Detail in Physically-based Real-time by John Dingliana and Carol O’Sullivan 8 for by Holly Rushmeier 44 Static Solution for Real Time Deformable Objects With Fluid Inside by Ivan F. Costa and Remis Balaniuk 9 Subdivision Surfaces in Geometric Modelling 46 Exploring Virtual Prototypes using Time-Critical by Pavel Chalmoviansky Rendering Techniques 10 A System for Out-of-Core Simplification of Huge Meshes by Enrico Gobbetti and Riccardo Scateni by Paolo Cignoni and Roberto Scopigno 47 Haptic Rendering of Molecular Properties 11 A Graphical System for Internal Density Modelling by Aleš Kfienek and Ludûk Matyska by Coons Bodies 48 The Virtual Museum of Daily Life in the 20th Century Bologna by Silvester Czanner and Roman Durikovic by Maria Chiara Liguori, Donatella Vasetti, Maria Elena Bonfigli, 12 Multiresolution Modelling of Surface and Volume Data Antonella Guidazzoli and Alessio Mauri by Leila De Floriani, Paola Magillo and Enrico Puppo 49 Augmented Reality Ð Seeing More 13 New CAD Tools for Aesthetic by Hellmuth Nordwig by Chiara Catalano, Bianca Falcidieno, Franca Giannini, 50 Interactive Computer Graphics at Rutherford Appleton Laboratory Maria Meirana and Marina Monti by Lakshmi Sastry, Gavin Doherty and Michael Wilson 15 -Based Acquisition of Virtual Gallery Model 52 Conceptual Free-Form Styling on the Responsive Workbench by Michal Haindl and Pavel Slavík by Gerold Wesche and Marc Droske 16 Computation of Medial Surfaces by Gábor Renner RESEARCH AND DEVELOPMENT

17 A Low-cost Optical 3D Scanner 53 Combining and Constraint Programming by Claudio Rocchini and Roberto Scopigno to Solve Real-Life Scheduling Problems by Philippe Baptiste 18 Shape Modelling and Processing: Methods and Applications by Bianca Falcidieno and Michela Spagnuolo 54 On-line Graph Coloring by JenŒ Lehel 19 Reverse Engineering Shapes by Tamás Várady 55 Boosting for Automated Text Categorization by Fabrizio Sebastiani, Alessandro Sperduti and Nicola Valdambrini 21 SOLEIL Ð Application of the Radiosity Method to the Physiological Simulation of Plant Growth 56 Pattern Formation in Electric Discharges by François Sillion and Cyril Soler by Ute Ebert 57 Design of Efficient Input/Output Intensive Applications 22 Image-Based Rendering for Industrial Applications by Ranieri Baraglia, Domenico Laforenza, Salvatore Orlando, by Samuel Boivin and André Gagalowicz Paolo Palmerini and Raffaele Perego 23 Colour Texture Modelling 58 Signware Ð A Learning Platform in Biology for Pupils with by Michal Haindl and Vojtûch Havlíãek Hearing Difficulties 24 Comparing Real & Synthetic Scenes by Kathrin Bolits using Human Judgements of Lightness by Ann McNamara EVENTS

26 Facial Repertoires for Animation on the Web 60 MFCS 2000 Ð Mathematical Foundations of Computer by Zsófia Ruttkay by Branislav Rovan 27 3D Medical Visualization: Breaking the Limits of Diagnostics 60 Joint ERCIM Workshop on and Environmental Modelling and Treatment by Brian Read and Thomas Lux by Andreas H. König and Eduard Gröller 61 EUROGRAPHICS 2000 Conference Report 29 Real-time Visualization through the Internet by David Duce by Xavier Cavin and François Cuny 62 AD2000 Ð The Third International Conference on Automatic 30 The Pond: Trawling the Information Sea Differentiation: From Simulation to Optimization by Lennart E. Fahlén, Jan Humble, Olov Ståhl, Jonas Söderberg by Uwe Naumann and Anders Wallberg 63 First Results of the Cross-Language Evaluation Forum 31 Advances in Computer-assisted Visualization by Carol Peters by Julian Gallop 63 FORTE/PSTV 2000 33 Graph Visualization of Complex Information by Tommaso Bolognesi by Ivan Herman 64 Announcements

35 Interactive Visualization of Protein Dynamics by Henk Huitema and Robert van Liere 67 IN BRIEF

2 Cover image: Round scenario with virtual objects, see article on page 42 ERCIM News No. 44, January 2001 KEYNOTE

uropean scientists should have access to national programmes in other member states. EThis is needed with a view to greater coherence in European research policy. Although no concrete target figures have been agreed, the political will to achieve this has been voiced by the ministers of research and technology policy in the member states, the European Commission and the Elly Plooij-van Gorsel, European Parliament. But more is required for the Member of the European Parliament and transition to the economy to be rapporteur on the European Research Area successfully achieved. E-mail: [email protected]

At present, it cannot be said that a uniform European research centres. ERCIM is a good example. But a European Research policy exists. Research policy of the member states and of the Area requires more than simple measures at the infrastructure Union are conducted alongside one another, without constituting . Infrastructures do not innovate, nor do electronic a coherent whole. Apart from the fragmentation of efforts, networks, although they must be in place. A European research isolation and compartmentalisation in national research systems, area can only come about if in addition to an advanced European member states invest significantly less in research infrastructure there is also a European identity and a European than the United States and Japan. These two factors mean that creativity. That means that member states must come to see the Union has simply come to further behind the United each other much more as partners rather than competitors. States in recent years. Opening up one another’s national programmes is a step in the right direction. As a result, the European research climate is declining rapidly. Research in Europe clearly needs a fresh boost. But more is called for. Innovations have led to society changing from an industrial society into a knowledge society. It is clear If the European Union is to invest in greater competitiveness that in future further development of this knowledge society and employment, a broader and more innovative approach is lies in prospect. That requires ongoing investment in education, required than has been the case to date. European member states training, but also investment paving the way for scientific should abandon their techno-nationalism and make greater breakthroughs. The rapid technological progress in the ICT efforts to achieve a European Union, at the research and sector is a major catalyst in this respect. technology level as well. Better coordination between the development of new The forthcoming enlargement of the Union makes this all the technologies and applications in the market is called for. Europe more necessary. The Framework Programme for Research and does not have any problem in converting Euros into research; Technology Development is a very useful instrument in what is troublesome is converting research into Euros. fostering international collaboration, but is has been found that Marketing, or adding knowledge to a product, appears to be this programme alone is insufficient to improve the collective more successful in the United States than in Europe. Increasing European research endeavours. More is required to actually knowledge is not an objective in itself. Converting that arrive at a European internal market for research. knowledge into innovation and industrial success is what we need to aim at. This is what creates employment and prosperity. The move made by Euro-Commissioner Busquin to create a I would like to make my contribution from the European context single European Research Area is something I very much and I am looking forward to any comments you may wish to welcome, speaking as rapporteur on the subject on behalf of make. the European Parliament. In the Guidelines for EU research activities (2000-2006), the Commission focuses primarily on optimising the infrastructure, by such means as linking up Centres of Excellence in a network and establishing virtual Elly Plooij-van Gorsel

ERCIM News No. 44, January 2001 3 JOINT ERCIM ACTIONS

Philippe Baptiste Winner of the 2000 Cor Baayen Award

The annual ERCIM Cor Baayen Award was presented the most promising young researcher in computer to Philippe Baptiste during a ceremony in Dublin on science and applied mathematics having completed 9 November 2000. The award is given every year to the PhD-thesis in one of the ‘ERCIM countries’.

The winner of the award, Philippe Philippe is also winner of the 2000 Prix country (a list with all national contacts Baptiste is today conducting research at Robert Faure, an award given every 3 can be found at http://www.ercim.org/ the French National Center for Scientific years to a French speaking researcher contacts/execom/). Research (CNRS) located at the under 35 by the French Operations University of Technology of Compiègne, Research Society. Nominees must have carried out their France. work in one of the ‘ERCIM countries’. Cor Baayen Award 2001 On the date of the nomination, nominees Philippe is 28 years old and has received The Cor Baayen Award for the most must have been awarded their PhD (or his PhD at the University of Technology promising researcher in equivalent) during the last two years. of Compiègne in 1998 while he was and applied mathematics was created in working for the research department of 1995 to honour the first ERCIM Each ERCIM institute is allowed to Bouygues, a diversified industrial group President. The Cor Baayen award is open nominate up to two persons from its leading activities in telecommunications, to any young researcher having completed country. A person can only be nominated , public utilities management, the PhD-thesis in one of the ‘ERCIM once for the Cor Baayen Award. The building/civil works, roads, etc. He has countries’, currently: Czech Republic, selection of the Cor Baayen award is the taken part into the development of Finland, France, Germany, Greece, responsibility of the ERCIM Executive optimization applications and of Hungary, Italy, Slovakia, Sweden, Committee. constraint programming libraries Switzerland, The Netherlands and the dedicated to the resolution of scheduling UK. The award consists of a cheque for How to nominate problems. Philippe has also been working 5000 Euro together with an award For proposing a nomination to your on complexity theory. He has fixed the certificate. The selected fellow will be national contact, fill out the Cor Baayen complexity status of several scheduling invited to the ERCIM meetings in Award Nomination Form available at the problems that had been opened for several autumn. ERCIM website. years. His work has been published in several international journals and he’s a Rules for Nomination Deadlines member of the editorial board of the Nominations for each country are made • 30 April 2001: nominations are to be Journal of Scheduling. See also his article by the corresponding ERCIM Executive received by the national contacts. ‘Combining Operations Research and Committee member (also referred to as • 15 May 2001: national contacts to send Constraint Programming to Solve Real- ‘national contact’). Those who wish that their selected two nominations to the Life Scheduling Problems’ on page 53. a particular candidate be nominated coordinator. should therefore contact the ERCIM Executive Committee member for their Further information can be obtained from your national contact or from the ERCIM Cor-Baayen Award coordinator Lubos Brim.

Links: The ERCIM Cor Baayen Award: http://www.ercim.org/activity/cor-baayen.html National contacts: http://www.ercim.org/contacts/execom/

Please contact: Lubos Brim Ð CRCIM Co-ordinator fot the Cor-Baayen Award E-mail: [email protected]

ERCIM President Gerard van Oortmerssen presents the 2000 ERCIM Cor Baayen award to Philippe Baptiste.

4 ERCIM News No. 44, January 2001 EUROPEAN SCENE

Strategic Workshops Ð Shaping future EU-NSF collaborations in Information Technologies

‘Bionics’, ‘Future Information Processing Technologies’ community following the call for for strategic and ‘Semantic Web’ are three of five strategic emerging workshops to be organized by ERCIM under the domains for a series of joint EU-NSF workshops. auspices of the European Commission and the US These three areas have been selected from over 350 National Science Foundation. different topics proposed by the European scientific

The European Commission (programme • high potential payoffs both in the EU The workshops are intended to facilitate IST-FET) and the US National Science and the US that make up for the long- brakethroughs in innovative domains, Foundation (CISE-NSF division) will term/high risk nature of research stimulate research activities and scientific organize a series of strategic research • existence of sufficient scientific and discussions of mutual interest. workshops to identify key research technological bases in both the US and Links: challenges and opportunities in the EU to entail balanced research http://www.ercim.org/EU-NSF/ Information Technologies. On the efforts European side, ERCIM solicited ideas Please contact: for high-level workshops from the ERCIM will organize a first series of Rémy Ronchaud Ð ERCIM office European IT scientific community. From workshops on the three selected topics in Tel: +33 4 92 38 50 12 E-mail: [email protected] the more than 350 suggestions, the 2001. The respective programme Strategic Workshop Review Committee committees will be nominated by the selected the following topics for joint Strategic Workshop Review Committee. future research initiatives: Participation is by invitation only, each • long-term/high risk nature of the workshop will be attended by a total of research involved, justifying risk 20 participants from both Europe and the sharing at international level US.

INRIA is growing at an Unprecedented Pace and is starting a Recruiting Drive on a European Scale

Last July, the French government decided to allocate by more than 50% in three years — a unique 425 extra positions to INRIA between now and 2003 opportunity to begin an ambitious policy and an as a means of bolstering the public research effort in unprecedented recruiting drive. and technology. INRIA will grow

The Institute’s strategy is based on the Through these five challenges and in academic or industrial experts. INRIA close combination of scientific closely coordinated interaction with other proposes also 200 opportunities: to achievement and technology transfer. In scientific disciplines, INRIA will focus prepare your PhD in one of its 100 order to meet its objectives, the Institute its efforts on some major fields of research project teams or to make post- intends to take up five major scientific applications, primarily telecommunications, doctoral stays of 12 to 18 months in one challenges: multimedia, health and biology. of its 100 scientific teams to develop a • mastering the digital infrastructure by research project. knowing how to code, compute and As early as 2001, the staff will grow by communicate over the Internet and over 180 personnel while the Institute’s The whole recruiting drive, job openings heterogeneous networks funding will follow suit with an increase and the projected time line for the • designing new applications that make of 60 million francs or 9.17 million euros. initiative are available on INRIA’s use of the Web and multimedia website. databases 115 civil service positions will be open Further information: • knowing how to produce reliable to French and foreign candidates: young http://www.inria.fr/ scientists, senior researchers, engineers • designing and mastering automatic and technicians. There will also be control for complex systems between 50 and 60 one- to three-year • combining simulation and . contracts for recent graduates as well as

ERCIM News No. 44, January 2001 5 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

Graphics and Visualization: Breaking New Frontiers

by Carol O’Sullivan and Roberto Scopigno

The two core issues in the past in overview of their group’s work in this computer graphics and visualization, have important field. Even if visualization been Modelling, ie capturing and technology has reached a rather mature representing the and state, research in the field is still very of virtual objects, and Rendering, which active with the aim of: extending the is the process of producing a realistic potentialities of the visualization image from such models. Although techniques, developing new approaches amazing advances have been made in for interacting with and navigating the both these domains, research is still data, and improving efficiency. The extremely active as the quest for true search for more - and time-efficient From the early graphical realism, and not just photorealism, solutions is still a basic goal, because the applications such as flight continues. Increasingly, extremely large impressive performance improvement of simulators, to today's stunning 3D datasets or models are required and systems is counterbalanced by special effects in movies, the level of accuracy, and hence the the increasing scale and complexity of the computer graphics have had a complexity, of such models is growing problems to be managed. significant impact upon the way dramatically. The article by Francois have been used to Sillion and Cyril Soler discusses a new The advent of low-cost 3D graphics represent and visualize the world. application of the radiosity rendering systems based on the PC There are many big problems left technique which allows accurate makes high-quality, real-time Interactive to be solved, some of which are of very large plant models. Graphics a ubiquitous and affordable reflected in the following pages Another for this explosion in resource. A powerful 3D graphics platform of this issue. model size is the improvement of the is now on the desk of anyone, from the devices for the rapid acquisition of 3D scientist to the student. New games digital models eg the non-invasive console hardware provides un-precented scanning devices widely used in medical support for real-time animation and and in industrial inspection, or rendering at a relatively low cost, bringing the optical 3D scanning devices which can cutting-edge 3D graphics into millions of be used to acquire the surface properties homes around the world. Many of the of real objects. The contribution by Holly advances in this area are based on the Rushmeier describes ongoing research results of pioneering graphics research. into 3D scanning systems at IBM Watson A big challenge in the future, is to capture Research. this technology, and to use it for other purposes, such as desktop VR for Visualization might also seem a rather education, and entertainment. For this consolidated field. Many techniques have to happen, real-time graphical systems been proposed in the last ten years for need to achieve levels of realistic image managing different data (surfaces, synthesis only achieved in the past with volumes, scattered points, vector fields, very computationally expensive rendering etc). Users have access to a wide range techniques such as path-tracing and of visualization modalities which allow radiosity. The animation of objects needs sophisticated and differentiated insights to be based on physical models, and of the data to be obtained. Consequently, entities in the world should be endowed visualization is a crucial with behaviours so that they can interact and analysis tool in the design and convincingly with the user. The simulation of many new products or performance/cost ratio may have systems. In the field of medicine, improved considerably, but such visualization has the potential to radically computational power is still not possible change the way that health is on low-end platforms. Therefore, administered, and significant progress has techniques which use multiresolution been made to date. Andreas Koenig and models, and levels of detail for all aspects Eduard Groeller from the Technical of simulation, are becoming increasingly University of Vienna have contributed an important.

6 ERCIM News No. 44, January 2001 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

CONTENTS

INTRODUCTION 29 Real-time Visualization 6 Graphics and Visualization: through the Internet Breaking new Frontiers by Xavier Cavin and François Cuny by Carol O’Sullivan 30 The Pond: Trawling the Information Sea and Roberto Scopigno by Lennart E. Fahlén, Jan Humble, Olov Ståhl, Jonas Söderberg MODELLING and Anders Wallberg New frontiers for visualization and 8 3D Scanning for Computer Graphics 31 Advances in Computer-assisted by Holly Rushmeier interactive 3D graphics are opened by the Visualization rise of Virtual Reality environments 9 Subdivision Surfaces in Geometric by Julian Gallop Modelling (CAVEs, workbenches, data walls) which 33 Graph Visualization of Complex by Pavel Chalmoviansky can greatly enhance the way that users Information 10 A System for Out-of-Core by Ivan Herman analyze data or objects, or navigate a Simplification of Huge Meshes 35 Interactive Visualization of Protein virtual world. Large tiled displays will be by Paolo Cignoni Dynamics and Roberto Scopigno available at a low cost in the near future, by Henk Huitema and Robert van Liere and applications requiring cooperative 11 A Graphical System for Internal 36 Physical Deforming Agents for Virtual Density Modelling by Coons Bodies analysis of data or navigation could Neurosurgery by Silvester Czanner benefit from these devices. Augmented by Michele Marini, Ovidio Salvetti, and Roman Durikovic reality allows graphical to be Sergio Di Bona and Ludovico 12 Multiresolution Modelling of Surface Lutzemberger superimposed on a real scene, either by and Volume Data 37 Visualization of Complex Dynamical projecting them onto real objects such as by Leila De Floriani, Paola Magillo Systems in Theoretical Physics and Enrico Puppo walls, or by viewing such virtual object by Anatoly Fomenko, Stanislav with see-through glasses. An exciting 13 New CAD Tools for Aesthetic Design Klimenko and Igor Nikitin by Chiara Catalano, Bianca Falcidieno, future development will be the 38 Simulation and Visualization of Franca Giannini, Maria Meirana Processes in Moving Granular Bed incorporation of real people into a virtual and Marina Monti world, as in the blue-c project described Gas Cleanup Filter 15 Image-Based Acquisition of Virtual by Pavel Slavík, František Hrdliãka by Markus Gross and Oliver Staadt in this Gallery Model and Ondfiej Kubelka by Michal Haindl and Pavel Slavík issue. 39 Watching Chromosomes during Cell 16 Computation of Medial Surfaces Division Europe is a primary player in the domain by Gábor Renner by Robert van Liere of Graphics and Visualization. This is 17 A Low-cost Optical 3D Scanner by Claudio Rocchini proved by the many research projects INTERACTIVE GRAPHICS and Roberto Scopigno 41 The blue-c Project active on these topics in European 18 Shape Modelling and Processing: by Markus Gross and Oliver Staadt laboratories, and ERCIM institutions Methods and Applications 42 Augmenting the Common Working support many of these research initiatives. by Bianca Falcidieno Environment by Virtual Objects by Wolfgang Broll The contributions collected for this and Michela Spagnuolo Special Issue give only a partial view of 19 Reverse Engineering Shapes 43 Levels of Detail in Physically-based by Tamás Várady Real-time Animation the many research projects active in this by John Dingliana and Carol O’Sullivan domain. Nevertheless, the spectrum of RENDERING 44 Static Solution for Real Time applications and topics covered by the 21SOLEIL Ð Application of the Radiosity Deformable Objects With Fluid Inside contributions is notably wide: basic Method to the Physiological by Ivan F. Costa and Remis Balaniuk Simulation of Plant Growth 46 Exploring Virtual Prototypes using modelling techniques (classical by François Sillion and Cyril Soler techniques, level of detail techniques, 3D Time-Critical Rendering Techniques 22 Image-Based Rendering for Industrial by Enrico Gobbetti scanning), rendering techniques (texture- Applications and Riccardo Scateni by Samuel Boivin and André Gagalowicz based graphics, image-based modelling 47 Haptic Rendering of Molecular of BRDF), classical animation and non- 23 Colour Texture Modelling Properties photorealistic rendering, physically-based by Michal Haindl and Vojtûch Havlíãek by Aleš Kfienek and Ludûk Matyska 24 Comparing Real & Synthetic Scenes simulation, virtual reality and haptic 48 The Virtual Museum of Daily Life using Human Judgements interfaces, augmented reality, 3D medical in the 20th Century Bologna of Lightness by Maria Chiara Liguori, Donatella imaging, molecular modelling, by Ann McNamara Vasetti, Maria Elena Bonfigli, information visualization, visualization 26 Facial Repertoires for Animation Antonella Guidazzoli and Alessio Mauri for simulation and training, and more. It on the Web 49 Augmented Reality Ð Seeing More is clear that this will remain a significant by Zsófia Ruttkay by Hellmuth Nordwig 50 Interactive Computer Graphics at area of activity for ERCIM institutes well VISUALIZATION Rutherford Appleton Laboratory into the future, and we hope that you gain 27 3D Medical Visualization: Breaking the Limits of Diagnostics and by Lakshmi Sastry, Gavin Doherty as much enjoyment from reading about and Michael Wilson these vibrant and exciting research Treatment by Andreas H. König and Eduard Gröller 52 Conceptual Free-Form Styling activities as we have while coordinating on the Responsive Workbench this special issue. by Gerold Wesche and Marc Droske

ERCIM News No. 44, January 2001 7 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

3D Scanning for Computer Graphics

by Holly Rushmeier

Three dimensional scanning has recently become a Watson Research we are developing scanning very active area in computer graphics. The systems for producing virtual objects that can be requirements for computer graphics are different from rendered with high visual quality. those of traditional scanning applications. At IBM

The capture of 3D representations of Our first major project using our scanning objects has been an area of research for system was creating a digital model of decades in , robotics and Michelangelo’s Florentine Pieta`. The metrology. Traditional uses for digital sculpture was scanned on site in the objects include industrial inspection, Museum of the Opera del Duomo in autonomous robot navigation and object Florence. The resulting model was used recognition. Each of these areas has by art historian Jack Wasserman to study different requirements and different issues such as how the work appeared approaches for data acquisition and after damage (now repaired) inflicted by processing. For inspection applications, the artist, and how it would have appeared metric accuracy is critical, and scanners in its originally intended site in a niche with great precision have been developed. above the artists tomb. Autonomous navigation applications, such as the use of robots in hazardous Many European research groups are environments, need to be very robust and Scanning Michelangelo’s Florentine Pieta. pursuing the development of inexpensive justify the use of expensive scanning scanners for computer graphics. Cultural equipment. Object recognition may use scanned objects can be used to populate heritage is a leading application for many both the shape and surface properties, but visually rich synthetic environments. of these projects. For example, the Visual only to the extent they are needed to These applications, while affecting a wide Computer Group at CNR Pisa, Italy, has compute a unique signature for the object. population, can only justify a modest developed a scanner based on a digital expenditure on individual scanners, camera and video projector. They have Unlike traditional applications, the end dictating the use of commodity recently used this scanner to capture a product of scanning for computer components in their construction. Also, model of a bronze Minerva in Florence. graphics is a model that can be used to these applications require interactive The VISICS group at the Katholieke render a realistic image of the object under display of the digital objects. This Universitat Leuven, Belgium has developed novel conditions, ie in a location or under requirement dictates that the output of the numerous capture techniques, including a lighting conditions that exist only in scanning processing pipeline must be a method for generating texture-mapped computer simulation. The emphasis is on form that can be rendered in real time. models from uncalibrated video streams. visual, rather than metric accuracy. A This method was used in reconstructing the systematic error in the shape may be less At IBM Watson Research, we have site of ancient Sagalassos in Turkey. important than errors in the color or focused on the development of scanning apparent shininess of the object. This systems that acquire surface properties as At IBM Research we are continuing to translates into a greater concern for well as shape, use commodity digital develop new scanning systems. We are estimating the spatial variation of the cameras, and produce texture-mapped exploring the use of alternative shape spectral bidirectional reflectance of a triangle meshes that can be efficiently capture systems based on digital cameras. surface, and less concern for precision rendered by graphics hardware. Our first We are exploring the trade-offs between range measurements. system was built around a commercial shape and texture resolution. We are also shape scanner that used projected light improving the processing pipeline to Applications for computer graphics stripes and multi-baseline stereo. We reduce user intervention to make scanning rendering of scanned objects include enhanced the scan with a novel accessible to a wide range of consumers. virtual museums, e-commerce and games. photometric system that acquired surface Links: In virtual museums, institutions can allow normals and albedo at a spatial resolution IBM Visual and Geometric Computing group: visitors to interact with virtual copies of 4 times the base geometry. We developed http://www.research.ibm.com/vgc objects that are physically too delicate to a processing pipeline for the acquired data touch, or possibly even to keep on that included new algorithms for meshing Please Contact: display. In e-commerce, a vendor may point clouds, for computing consistent Holly Rushmeier Ð IBM Watson Research Tel: +1 914 784 7252 offer a potential consumer the capability normals from photometric data, and for E-mail: [email protected] to view an object in a variety of the alignment and integration of surface customized configurations. In games, texture .

8 ERCIM News No. 44, January 2001 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

Subdivision Surfaces in Geometric Modelling by Pavel Chalmoviansky

Scientists at Comenius University, Bratislava, are manipulation of the geometric properties of shapes, a developing techniques for efficient and robust task that has been particularly problematic in the past.

Construction using subdivision surfaces figure below, the whole process of the required. Modelling inside CAD systems is a technique in geometric modelling subdivision can be seen. Usually several in this area produces more problems: which combines both computational iterations are enough for high accurate during the intersection of two surfaces we simplicity and ease of control. Although results. create objects that are not primarily there are some unsolved problems, this products of the subdivision scheme. The appears to be a growing trend in recent There are several known schemes of this solution for this situation should be years. After the era of mainly parametric, type published by Loop, Catmull-Clark, compatible with the whole modelling implicit and functional modelling, there Doo-Sabin, Kobbelt and others. All of system. Mainly in CAD systems, high is now a method which offers more them work on an almost regular cell order continuity surfaces are required. flexible (than in the case of implicit complex with triangular or quadrilateral The known methods suffer from huge

Subdivision process. surfaces) and more intuitive (than in the faces. Final surfaces can interpolate the numbers of vertices involved in the case of integral or rational surfaces) values of p(i) at vertices and one can also computation of the next iteration. control and modelling of the final shape. mark special edges which play the role of Moreover, the final surface can be easily boundaries during the construction and A wide range of unsolved questions arise represented as a multiresolution model. thus other rules are applied upon them. if we consider also irregular cell There are also so-called combined complexes. The set of solutions is very The first subdivision method was schemes which take special care at the sparse here. Subdivision surfaces were developed more than 20 years ago, but boundaries of the surface. The schemes also used in real applications. In 1997 the the method achieved more popularity only produce surfaces of lower degree (if they first animation using only subdivision in the 90’s. During these years the are polynomials) and lower order of surfaces was made by Pixar - Geri’s mathematical foundations were laid and continuity. This is however not enough Game, and Toy Story II also used them. several problems (eg of continuity and for some applications and further schemes Please contact: parameterisation) were solved. However, are sought. Pavel Chalmoviansky Ð SRCIM/ success was only achieved with static Comenius University subdivision methods. The main open problems are to be found Tel: +421 7 6029 5229 when we consider the manipulation of E-mail: [email protected] The principle is to refine the structure of these objects. As the number of cells the surface representation up to a state increases exponentially, it would not be with which we are content. We start with possible to maintain a high number of cell complex K(0) and function p(0) on objects at their full representation. the vertices of K(0). The refinement or Adaptive methods represent objects using subdivision has two phases: We construct several layers. The more layers one K(i+1) from K(i) and p(i+1) from p(i). displays, the more details one can see. The rules are usually local and simple, The cell that will be cancelled in a layer hence the new complex and the new can be computed according to variational function can be easily computed. In the calculus. More efficient methods are still

ERCIM News No. 44, January 2001 9 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

A System for Out-of-Core Simplification of Huge Meshes

by Paolo Cignoni and Roberto Scopigno

Recent evolution in graphics technology has been and implementation of a for the out-of-core impressive, and the management of very complex management of huge 3D meshes. One of the main models is now possible on inexpensive platforms. We goals is to support high quality simplification of huge present the results of a project aimed at the design meshes.

The performance of graphics subsystems has improved enormously in the last couple of years, but unfortunately the complexity of 3D scenes and graphics applications has also increased greatly. A very critical aspect in interactive 3D graphics is the complexity, in terms of number of triangles, of the meshes to be managed. Huge meshes can be easily produced in many applications: by fitting isosurfaces on large volume datasets, by converting surfaces into triangulated meshes, by 3D scanning real objects. All 3D scanners perform a regular sampling on the surface of an object, returning triangle meshes of complexity directly proportional to the scanner’s sampling resolution and the object surface area. For example, the size of the S. Matthew 3D model, recently produced by the Digital Figure 1: An image from the simplified digital Michelangelo Project, is around 370M S. Matthew mesh: a portion of the high resolution mesh is top-left (4M triangles), triangles. Obviously, these complex a simplified intermediate LOD is top-right meshes introduce severe overheads in (48K tr.); a drastically simplified mesh transmission, rendering, processing and (500 tr.) is shown bottom-right, with shape storage. Mesh simplification and level- detail preserved by a resampled texture in bottom-left image. Figure 2: Images from a simplified digital of-detail (LOD) management can in many David mesh (only 215 K triangles). cases efficiently reduce these overheads but only very large and costly systems can manage meshes characterized by such of a new OOC simplifier, which adopts platform with limited memory. No a huge size: RAM size is often a severe a simplification approach based on edge existing systems support this bottleneck. collapse. functionality. The hierarchical OEMM data structure is at the base of our mesh The management of complex geometric Management and Simplification management and simplification system. datasets is a very active research area. A of Huge Meshes The OEMM structure supports dynamic number of sophisticated technologies We briefly introduce a system which ‘on request’ loading of portions of the have been developed to simplify the supports huge mesh management current dataset from secondary memory, geometry while maintaining accuracy and (including mesh editing, visualization, for the browsing or updating of preserving any attributes contained (eg simplification and detail preservation) on geometrical and topological information. colour or pictorial data), and to use LOD PC-based platforms. The system uses a This data structure can be constructed techniques to speedup rendering. innovative hierarchical data structure, starting from a generic raw triangle-based called Octree-based External Memory mesh representation. The following We have recently studied a problem Mesh (OEMM), to maintain the data on modules can be built on the top of this which arises while 3D scanning large external memory, to load only selected data structure: artefacts: how can simplification sections dynamically in main memory, • a mesh editing library that allows the technology be extended to manage and to preserve data consistency during user to perform simple editing actions datasets that do not fit the main memory local updates. (such as 3D scanning and rapid of most computers? Our aim is to design prototyping). These include topological an Out Of Core (OOC) mesh simplifier. Our goal has been the management of check of the mesh, detection of non- Here below we describe the architecture really huge meshes on a PC-based manifold components, detection of

10 ERCIM News No. 44, January 2001 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

holes and automatic or user assisted external memory management is Michelangelo Project) and the hole-triangulation, elimination on affordable. improvement of shape accuracy supported request of complex vertices and faces, by our solution was significant. elimination on request of small The only other approach which supports components, etc out-of-core simplification of huge meshes This project has been developed with the • a mesh simplification library, which is the out-of-core clustering solution financial support of the Progetto performs out-of-core simplification recently proposed by Lindstrom at Finalizzato ‘Beni Culturali’ of the Italian adopting the quadric error metric Siggraph 2000. This solution is very easy National Research Council and the approach (see Figures 1 and 2) to implement but, unfortunately, it has the Project ‘RIS+’ of the Tuscany Regional • a detail preserving library, which drawbacks of the clustering approach: the Government. Scanned data are courtesy resamples bump- or rgb-texture to simplified mesh quality is much lower of the Digital Michelangelo Project, encode the high frequency detail lost than the quality of simplified meshes Stanford University (USA). during simplification (see Figure 2) produced with the edge collapse Links: • a visualization module and GUI approach, simplification is not adaptive Group CNR-PISA: components to surface shape and many topological http://vcg.iei.pi.cnr.it degenerated situations are introduced. We Our system supports management and have compared the results produced by Please Contact: simplification of up to O (109) triangles OOC Clustering and our OEMM edge Paolo Cignoni Ð IEI-CNR Tel: +39 050 315 2927 meshes. The time overhead due to the collapse simplifier on the same dataset E-mail: [email protected] (the David mesh of the Digital

A Graphical System for Internal Density Modelling by Coons Bodies by Silvester Czanner and Roman Durikovic

The rise of solid modelling as a principal medium for information neededfor some engineering . mechanical product description can be traced to the Many unambiguous solid representation techniques requirement of informational completeness of like primitive instancing, cell decomposition, geometric representations. Unfortunately, traditional constructive solid geometry have one limitation, they geometry-based systems do not contain the important cannot offer ways of representing internal behavior.

In the mechanical industry it is very an arbitrary 3D shape. The elements used structure. The simplest and most limited important to know not only the are so called parametric solids. We are is Coons body 0 where only the shape of description of the surface, but also about using three types of parametric solids, quadrilateral can be changed. Coons body the interior and the density of the object. namely Coons body 0, 1 and 2. The 1 allows the user to control the interior Therefore, there was a need for a system bodies are defined as an extension into by modification the control points along which makes it possible to model the four-dimensional space of well-known the edge curves of a parametric solid. The interior density of the 3D parameter Coons patches in B-splines form. most general is the Coons body 2, having solids. the ability to modify shape and interior Each of the three Coons bodies have by all control points within the boundary A Project between the Software different abilities to control the interior surfaces of a parametric solid. Department of the University of Aizu and the Department of Computer Graphics To understand the interior structure of 3D and Image processing of the Comenius objects several visualization methods are University exists to support and develop used. The easiest way to visualize a a new methodology for representing an parametric solid is to display it as a set of internal structure of density of 3D objects. isoparametric surfaces. Another option is The main goal of the project is to develop to visualize a shaded object together with a graphical system for modelling and projected parametric curves visualizing the interior structures and corresponding to constants u, v and w. density of 3D objects. It allows a The most difficult option is to show the developer to make modifications into interior changes in time as internal structure of 3D object in a natural metamorphoses between two Coons and intuitive way. bodies. Figure 1: Nail as Coons body 2 with two different interior density distributions applied The system can also modify and as initial conditions for two FEM simulations The most useful elements are Coons composite 3D objects (basic elements) to of internal tension. bodies 2. Since, any of them is a B-spline

ERCIM News No. 44, January 2001 11 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

Figure 2: Modelling a spool as Coons body 2 starting from a cube. Figure 3: Shape represented with five Coons bodies 2..

volume, the shape continuity is simply The authors see the advantage of this Please contact: controlled in a similar way as it is for B- graphical system in the ability to define Silvester Czanner Ð SRCIM / splines. The system can create more the boundary and internal initial Comenius University Bratislava complicated shapes such as a composition conditions prior to numerical simulations. E-mail: [email protected] of several Coons body elements. Parametric solids can be effectively Roman Durikovic Ð University of Aizu, Japan applied on finite elements with irregular E-mail: [email protected] Implementation is done under the Linux or trimmed boundaries to avoid difficult Red Hat 6.2 and the problems in model design for Fine objects are stored in a scientific format Elements Methods. SILO.

Multiresolution Modelling of Surface and Volume Data

by Leila De Floriani, Paola Magillo and Enrico Puppo

Multiresolution models have become very popular graphics group at the Department of Computer and during the last few years to manage the increasing Information of the University of Genova. complexity of surface and volume . Major applications concern terrain modelling in the This topic is the subject of several research projects context of Geographical Information Systems, 3D carried out by the geometric modelling and computer object modelling, and volume visualization.

The resolution of a geometric model is The key is that usually high Since the production of a model at a related to the number of primitives used resolution is not needed everywhere; the specified (variable) resolution is to represent the details of a shape. level of resolution can be variable in space computationally expensive, a Resolution has an impact both on the and time. For instance, a three- multiresolution model is built off-line as accuracy of the representation (thus, on the dimensional scene can be rendered using a comprehensive structure which similarity of the model to the real object) a resolution which is higher near the organises several pre-computed and on the size of the data structures (thus, viewpoint, since many details are required resolutions and supports a fast on-line on computational efficiency). only close to the observer; in an combination of different resolutions into interactive visualization program with a a single model, according to user Modern tools both for acquisition of moving viewpoint, the resolution should parameters. geometry from real world, and for be adapted dynamically. computer-aided design, can provide The DISI research group has developed models at very high resolution. These Variable resolution is crucial, eg, for real- the Multi-Tesselation (MT) model, a models are very accurate but often too time visualization in virtual reality general multiresolution model for spatial large to be effectively used in rendering, (aircrafts, ships, buildings...), scientific data that does not depend on: especially at an interactive rate. data visualization (three-dimensional • the dimension of the data (it can model Multiresolution refers to the possibility fields such as temperature, pressure...), surfaces in space, solid objects, volume of representing a spatial object at different geographic information systems (terrain), data, scalar fields in any dimension...) resolutions, trying to balance the opposing medicine (volume data from TAC, • the process used to generate the issues of having sufficient accuracy and SPECT...). resolutions that contribute to the model small size.

12 ERCIM News No. 44, January 2001 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

Using a variable resolution in the view frustum for terrain rendering allows high quality images with reduced computational effort. Here, resolution (ie size of triangles) decreases with distance from the viewpoint (see top view), while the size of projected triangles is constant through the whole image (see view).

(unlike many other models which must We have recently released a software render huge three-dimensional data sets be built only with a certain ) library implementing the MT model (see thanks to a multiresolution approach. • the criteria used to extract representations http://www.disi.unige.it/person/MagilloP/ Links: at variable resolution from the MT (the MT/), which is intended as a general tool Geometric Modelling and Computer user can constrain either the granularity for supporting variable resolution in Graphics group, University of Genova: or the accuracy on a local basis by geometric applications. http://www.disi.unige.it/research/ providing reference entities such as areas Geometric_modeling/ of interest, viewpoints...). MT has been used as a basis for: Please Contact: VARIANT, a prototype system for terrain Paola Magillo - DISI University of Genova The MT model can be specialised for modelling, which supports variable Tel: +39 010 353 6705 specific types of data, construction resolution; and TAn2, a system for E-mail: [email protected] processes and/or operations, leading to developed in very efficient implementations in those collaboration with the Visual Computing specific cases. Group at CNR in Pisa (Italy), which can

New CAD Tools for Aesthetic Design by Chiara Catalano, Bianca Falcidieno, Franca Giannini, Maria Meirana and Marina Monti

The definition of tools to support design intent of light over the surface. In order to provide tools for representation and management during product the direct manipulation of the shape through character development is a key issue for new CAD systems. In curves, in the FIORES project the form feature concept the context of product styling development, the design has been extended to the aesthetic design domain. intent is represented by the shape of the resulting Algorithms for surface modification through target product and can be schematized by the defining shadow line specification have also been character curves and by the behavior of the reflection implemented.

In a competitive world wide market, where rendering and animation facilities for the (eg NURBS/B-Splines control points and product life time is being reduced, quality purposes of simulation and evaluation, thus nodes) completely unrelated to the design and cost are not necessarily the only reducing the number of physical prototypes intent of the stylist. important key elements for success: the and cutting design costs. aesthetic aspect is also becoming In order to overcome these limitations, fundamental in customers’ decisions. Unfortunately, current CAD/CAS systems new modelling tools are being developed. Furthermore, the availability of new do not support the fast generation of The Brite-Euram Project FIORES materials and production tools makes the alternative shapes starting from existing (Formalization and Integration of an fabrication of very complex shapes possible, models. It is sometimes very difficult to Optimized Reverse Engineering Styling thus permitting a greater freedom in design. change a small detail and can cost more Workflow) focused on this problem. The than re-creating the entire model.This is two main aspects regarding styling Computer-aided tools have been developed due to the that the modelling activity considered by the project are to support styling, providing high quality is based on low level geometric elements identification of free form features for

ERCIM News No. 44, January 2001 13 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

Figure 1: An example of a car having very evident character lines.

shape modification, and engineering in reverse techniques for surface refinement.

Free Form Features Feature technology is employed to speed up the model creation process and permit the reuse of existing models. Features are well known and common in mechanical Figure 2: A snapshot of the step-like feature implemented in the FIORES prototype system by Dassault Systemes and University of Kaiserslautern. design and their application to styling has required an analysis of the various elements characterizing the product shape. features. Two main classes of features to special curvature lines. The have been identified: those corresponding then modifies the object surfaces to make We have identified two relevant to elastic deformations and those the derived evaluation lines look good. categories of form features: regarding elimination of surface regions. • Structural features, consisting of key A feature taxonomy has been specified With the engineering in reverse (EiR) lines for the overall shape definition. on the basis of these two classes. It is approach adopted by FIORES and Examples of structural features are aimed at allowing the identification of a implemented by Matra Datavision contours, object profiles and sections, useful set of intuitive parameters (possibly together with UDK Utveckling and or structural character lines having an with aesthetic value) and conditions, Formtech, the modification loop can be overall aesthetic impact. which can be hard-coded, to define avoided: the designer can directly specify • Detail features, consisting of local shapes meaningful deformation with predictable a preferable target evaluation line, and added to the object surface for aesthetic behavior. A feature class has been insert additional constraints (such as the and functional and to enhance implemented to demonstrate the modifiable area). The system will return the visual effects of character lines. feasibility of this approach. a surface with the given evaluation line. An optimization approach is used to We have mainly focused on detail Engineering in Reverse compute the best solution, ie the solution features and have proposed a formal This approach aims at shortening the that according to a given criterion is as definition of feature types on the basis of frustrating loop reiterating surface close as possible to the desired target. detail features obtained by modifying a modification and aesthetic evaluation of Links: given free form surface locally. the product model. Many industrial FIORES web page: http://www.fiores.com products, such as cars, are evaluated in This classification is based on the terms of certain evaluation lines that give Please contact topological and morphological the expert designer an idea of the aesthetic Bianca Falcidieno Ð IMC-CNR characteristics associated with the quality of the object. These lines range Tel: +39 010 64751 E-mail: bianca(franca)@ima.ge.cnr.it deformation functions used to create the from section curves and reflection lines

Figure 3: A surface with ‘bad’ highlights. Figure 4: Specification of the target highlight (B) for correcting the wrong one (A) and the one (C) resulting from the application of the engineering in reverse method.

14 ERCIM News No. 44, January 2001 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

Image-Based Acquisition of Virtual Gallery Model by Michal Haindl and Pavel Slavík

The purpose of this Virtual Reality project was to build of Modern Art of the National Gallery in Prague is a virtual gallery model from real image data intended to serve as the core of the web based measurements utilising results from the VIRTUOUS National Gallery . project. The resulted virtual model of the Department

Customary manual creation of virtual solution to this problem is an reality models of real world scenes is automatic path generation that defines tedious and error-prone work as the a trajectory of the virtual walk scene complexity increases and any through. The path parameters (entry automation can substantially reduce and exit points, obstacle avoidance - the laboriousness and consequently eg, usage of a staircase for the cost of the whole process. Visual handicapped visitors, etc.) are defined sensors offer the ideal source of data by the user. The process of the path for such a task especially when range finding is in principle finding a path and vision sensors are already in a underlying graph describing the common and their mutual registration virtual scene and constrained by can be done using either standard requested art works to be seen on the photogrammetric techniques or an Part of the exhibition in the Department of Modern Art. way. appropriate sensor setup. (halls, corridors, etc.) containing mostly The virtual model of the Department of To make virtual worlds realistic detailed planar-faced objects. However although Modern Art of the National Gallery in scene models must be built. Satisfactory even if a single planar face is correctly Prague was semi-automatically created models require not only complex 3D found, its corresponding range pixels in the VRML2 language so it can be shapes accorded with the captured scene, values are too noisy to be of direct use for easily placed on the web server with but also lifelike colour and texture. the face model construction. Such a model sufficient resources. The model recently Textures provide useful cues to a subject would have many small noise-generated describes interior public exhibition areas navigating in such a VR environment, and faces which have to be eliminated. Noise of the gallery building together with the they also aid in the accurate detailed filtering is done through the least-square exhibition panels holding paintings from reconstruction of the environment. This fitting of a plane to a detected face the gallery art collection and can be used increases significantly the realism of the perimeter pixels and only the rectified not only as the visitors virtual information synthetic generated scene. face pixels are subsequently used for a system but also for gallery workers to virtual model face geometry inference. model optimal show arrangement for Our pilot system for the virtual model exhibitions to be prepared. Although acquisition currently restricted to planar- Thematic maps from the range segmentor recent technology advances already faced scene objects was used to build the serve also as the region maps for the enables automatic or at least semi- gallery model. The core part of this VR registered colour textured images. If a automatic construction of distributed modelling system is a fast range image detected face has sufficient area for a virtual models, further research is still segmentation algorithm for planar-faced reliable texture model inference then this needed especially in the area of general scene objects. The Department of Modern model is inferred from the corresponding surface object modelling. Similarly the Art, housing a collection of images, projected colour textured image area, VRML2 language has many functional drawings and statues from the period of otherwise the face is classified as too restrictions and some better distributed 20th century, is located in a functionalistic small for individual texturing and the virtual reality modelling language is building in Prague, where the planar- texture model can be assigned clearly required. faced object restriction is quite acceptable. interactively if required. Textures are Links: The segmentor combines a discontinuity modelled using our fast multi-grid VIRTUOUS project: detection step and the subsequent line Markov random field based model and http://www.ee.surrey.ac.uk/EE/VSSP/ based region growing guided by these combined with model geometry 3DVision/virtuous/virtuous.html detected discontinuities. Range information. National Gallery demo: segmentation results subsequently http://www.utia.cas.cz/user_data/haindl/ PR/demHP.html supervise the following - geometry Resulting gallery model is quite complex inference and texture analysis steps. hence it has to be completed with a virtual Please contact: environment navigation system. Michal Haindl Ð CRCIM (UTIA) The gallery model is composed from Otherwise its user can be easily lost in the Tel: +420 2 6605 2350 mutually connected partial submodels labyrinth of gallery virtual corridors. Our E-mail: [email protected]

ERCIM News No. 44, January 2001 15 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

Computation of Medial Surfaces

by Gábor Renner

Medial surfaces as skeletons of three-dimensional Modelling Laboratory of SZTAKI fast and robust objects are geometric abstractions of solids, that methods have been developed to calculate the medial facilitate volumetric and shape reasoning in many surfaces of solids starting from their boundary areas of engineering computations. In the Geometric representation.

The medial surface or medial axis Part transform is a skeleton of a three and medial dimensional object, which is defined as surface. the locus of the centres of the interior spheres with maximal radius. Together with the radius of the maximal spheres as a scalar function it reflects the topological and geometrical features of an object in a very compact form. Moreover this representation is unique for each ‘valid’ them. This is mainly due to the concluded that in a great majority of (ie manifold) solid object. It can be computational complexity inherent in geometric situations a solution can be invertible, and the object can be medial surface computations based on the computed with analytic tools. In the reconstructed from the skeleton. usual boundary representation of real remaining cases, specific geometric three-dimensional objects. As a situations (that occur frequently with The boundary representation of a solid - consequence, to develop a robust, technical objects) can be detected, where which is used in commercial CAD/CAM efficient and at the same time user friendly again analytic solution is possible. Thus systems - builds up a complete topological algorithm is very important. A realistic the use of a nonlinear system solver can structure of the object, defining the simplification of the starting geometric considerably be reduced which results in connectivity relationships between the model or the skeleton to be computed – fast and stable computations. faces, edges and vertices. These are or both – is generally needed in order to associated with corresponding geometric overcome the tremendous amount of To illustrate the efficiency of the above entities describing their shape. While computation and to obtain results with case analysis a part with the corresponding many practical engineering tasks can be acceptable computational costs. medial surface is shown on the figure. The solved using this mechanism, there are object was approximated by 94 facets, and important questions which can only be In our research we concentrate on the the total number of skeleton vertex answered using very cumbersome and medial surface computation of faceted calculations was 121925, out of which time-consuming procedures. Such topics objects. This means that curved faces of only 305 (0.25%) required nonlinear are, eg to recover proximity relations of the object are approximated by planar system solving. some parts of the object or to obtain an facets and curved edges by several overall view of the object. Based on the straight lines. This kind of approximation The speed and stability of medial surface skeleton these questions can be answered can easily be generated automatically in calculations can be greatly increased by easily, and a number of geometric many solid modelers. We have developed special topological considerations. Our operations (eg proximity calculations, an algorithm, which is based on the dual research concentrated on the elimination shape decomposition) can be performed structure of the skeleton, and determines of redundant calculations by analysing very efficiently. This is the reason why the spatial location of the vertices of the the structure of the object in a skeletons can be used in many skeleton, together with their topological preprocessing step (by applying multiple engineering applications with a great connections. One of our main start-points), and the development of a benefit. Such areas are for example NC improvements is the increase of speed and new algorithm to reduce the size of the process planning, design of robot robustness of geometric calculations. problem by subdivision (divide-and- trajectories, recognition and creation of conquer algorithm). geometric features, subdivision of a solid To compute the vertex positions of the into simpler parts for example for skeleton, a highly nonlinear system of Development of programs and methods automatic finite element mesh generation. equations must generally be solved. For is continuing in collaboration with the the numerical solution of the system a Ecole Polytechnique Fédérale in In spite of the above advantages of medial general nonlinear solver may be used, Lausanne, Switzerland. surfaces, there is no commercial which proves to be a slow and unstable Please contact: CAD/CAM or geometric modelling process. We have carefully analysed the Gábor Renner Ð SZTAKI system, which is based on the skeletons system of equations from analytic and Tel: 36 1 3868782 or at least supports the calculation of geometric points of view. It can be E-mail: [email protected]

16 ERCIM News No. 44, January 2001 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

A Low-cost Optical 3D Scanner by Claudio Rocchini and Roberto Scopigno

Automatic 3D acquisition devices (often called 3D include: high accuracy, affordable costs of the scanners) can be employed to build highly accurate scanning device and an improved of the data models of real 3D objects in a cost- and time-effective produced. We present the characteristics of a new manner. The Visual Computing Group of CNUCE/IEI- low-cost 3D scanner, based on structured light, and CNR has experimented this technology in a particular discuss the first results of its use in a project for the application context: the acquisition of Cultural 3D acquisition of a bronze statue. Heritage artefacts. Specific needs of this domain

Cultural Heritage is one of the few fields acquisition of a section of the Coliseum • use only consumer technology, to ensure where classical 3D modelling tools result in Rome by an Italian research team. affordable hardware costs and enable inadequate to model the shape of the easy technological improvements objects or artefacts of interest. This is both 3D scanning technology has evolved in • support sufficient accuracy and due to the shape complexity of most the last few years, but unfortunately not resolution, ie midway between artefacts (eg sculptures) and also to the as fast as other electronic devices. One commercial low cost laser scanners and high accuracy requested. The 3D model critical problem is the high cost of the high quality ones in many cases should not only look scanning devices. A good scanner, able • ensure easy operability and flexibility. visually similar to the real object, but to produce data at the accuracy required should also be very accurate, from a by this class of applications, often costs The scanner has been designed around geometrical point of view. This is more than 100K US$. This discourages two very common electronic devices: necessary for many applications such as the use of this technology in many • a video projector (preferably DLP the construction of 3D catalogues, the Cultural Heritage institutions, especially technology), to project structured light automatic reproduction of copies, the use in our national context. patterns on the object to be scanned of 3D models in the context of restoration • a digital still camera, to acquire images plans, etc. We have thus studied the design of low of the object under structured lighting. cost and medium quality optical scanners. 3D scanners can give a very precise The result is the design and prototypal Both devices are driven by a software reconstruction of the shape of a real construction of a new scanner based on module running on a standard PC. This object. This technology has been adopted consumer electronic technology. The acquisition module produces a series of in a number of recent cultural heritage architecture and features are described patterns (stripes with decreasing width, projects. We can cite the Digital briefly here. projected by the video projector) and Michelangelo Project of Stanford drives the camera. Photos are taken to University, the acquisition of An Inexpensive 3D Scanner acquire: images of the distorted patterns Michelangelo’s Pietà in Florence by the Our scanner has been designed to fulfil (from which the geometry is IBM T.J. Watson Laboratory, or the the following goals: reconstructed, producing a range ), and images of the object under different

Figure 1: The low cost scanner developed at CNR, which uses structured light and consumer electronic Figure 2: A preliminary reconstruction of the technology. head of the Minerva.

ERCIM News No. 44, January 2001 17 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

illumination conditions (from which the As usual in 3D scanning, complex objects This project has been developed in illumination-invariant color, or ‘albedo’, are modelled via the acquisition of a set cooperation with the Centro di Restauro of the object surface is reconstructed). of partially overlapping range scans. Our (Restoration Laboratory) of the The color images by definition are self- is used to register Soprintendenza Archeologica Toscana in recorded with range maps. and merge all the range maps in a single Florence, and with the financial support triangulated mesh. of the Progetto Finalizzato ‘Beni The scanning system depicted in Figure 1 Culturali’ of the Italian National Research produces: The scanner is currently in use at the Council and the Project ‘RIS+’ of the • a range map of each selected surface Soprintendenza Archeologica Toscana, Tuscany Regional Government. portion, with sample density of 0.7mm in the restoration of the Minerva, a bronze Links: (using a 1024x768 video projector, statue of the Museo Archeologico of Visual Computing Group CNR-PISA: located at approximately 1.3 m from Florence. The entire statue was scanned http://vcg.iei.pi.cnr.it the artefact) in October 2000, producing 146 range • an illumination-invariant colour texture maps and around 3000 images. We are Please Contact: (aligned with the corresponding range currently processing this data, and a Claudio Rocchini Ð IEI-CNR Tel: +39 050 315 2926 map) which represents the pictorial complete model will be ready at the end E-mail: [email protected] detail present on the acquired surface of 2000. Preliminary results are shown in section. Figure 2.

Shape Modelling and Processing: Methods and Applications

by Bianca Falcidieno and Michela Spagnuolo

Shape modelling is both a cognitive and creative Computer Vision. The Computer Graphics Group at process and reasoning about shape is a common way the Istituto per la Matematica Applicata (IMA-CNR), of describing and representing objects in engineering, Genoa, has been working on different aspects of architecture, medicine, biology, as well as in daily life. shape-based modelling since 1980. The focus of the Computers have revolutionised the field of shape research has been on the definition of abstraction modelling and opened up new applications, such as tools for deriving high-level descriptions of complex Computer Aided Design, Computer Graphics and shape models.

A key objective in research on shape- representations are necessary for their model and its features, when we use based modelling is to enhance traditional approximation function, on the other, it abstraction and/or classification tools to approaches to surface modelling through is desirable to communicate with the group low-level geometric entities into the definition of representations that are model using higher-level abstractions, classes with an associated meaning closer to our semantic coding of shape directly connected with specific properties (conceptual model). perception. Surface reconstruction of the shape. For example, when techniques can be used to faithfully describing a landscape in words we would IMA-CNR is working in a number of reproduce complex shape models of use terms such as ‘mountainous’, or projects using shape-based reasoning for existing or virtual objects, which may be expressions such as ‘many ridges and shape representation and processing. thought of as ‘digital mock-ups’ to be ravines’. It is thus clearly not important Reverse engineering is a good example visualized or analyzed using appropriate to communicate the number and the shape of an interesting application field for our software. If, on the one hand, geometric of the polygons that compose the digital approach. Digitizing machines usually

a) b)

Figure 1: A digitized object with the structural Figure 2: An object with the colour-coded characterization of its critical levels (a) and the features recognized. corresponding shape graph (b).

18 ERCIM News No. 44, January 2001 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

perform sampling along predefined at IMA concerns the extraction of directions, sections or profiles, which morphologic features from surfaces induce at least a partial ordering in the represented by cross sections. Cross data set. Shape-based techniques are sections implicitly contain information especially efficient in this context since about features like ridges, ravines, peaks, structural features of the underlying etc., which can be automatically surface can be easily detected by recognized in order to improve the analysing the configuration of points accuracy of the reconstructed model. along the sampling directions. Reconstruction can thus be performed a Finally, we are also studying the use of posteriori with respect to the recognition digital object models for the visualization of a rough shape skeleton. This is shown and coding of uncertain information. in Figure 1, where the approximating Figure 3: An object defined by uncertain Different forms of uncertainty and errors triangulation is fitted to the recognized sample points. can be introduced during the modelling structural features. IMA has also signed pipeline as data are acquired, modelled an important research contract with and visualized. Different approaches for Monolith Ltd, Tokyo, within the and these properties must be defined coding uncertainty have been considered, framework of a programme funded by formally so that they can be easily eg intervals, distribution private and public Japanese research associated with a geometric representation. functions and fuzzy numbers. Graphical organizations, for the development of a primitives, such as cubes and spheres, can terrain modelling system based on the Understood mathematically, topology is be used to visualize sets of spatial topological and semantic structure of the study of shape properties that do not measures affected by uncertainty, while natural surfaces. change under deformation; it provides a animation and other visualization sound theoretical basis for the techniques such as transparency, can Our main current research efforts are in formalisation of a number of problems effectively represent the surface at various the field of computational topology, related to shape and its understanding. levels of presumption (see Figure 3). incorporating abstraction mechanisms in For example, we are currently Links: shape modelling. To properly represent investigating the use of a critical point http.//www.ima.ge.cnr.it shape information, automatic mechanisms configuration graph, automatically must be developed to associate properties extracted from a triangulated object Please Contact: (eg meaning) with geometric data within model, to define a sketch of the shape Bianca Falcidieno or Michela Spagnuolo Ð the object description. To do this, suitable for classification and IMA-CNR Tel: +39 010 64751 properties, or shape features, are identified compression purposes (see Figure 2). E-mail: [email protected] in order to describe objects effectively, Another technique studied and developed

Reverse Engineering Shapes by Tamás Várady

In many areas of engineering, medical sciences or art realistic expectation in the geometric modelling there is a strong demand to create an appropriate community to generate accurate and topologically computer representation of existing objects from huge consistent models, ready to be used in CAD/CAM. sets of measured data points. This process is called Recent research results of the Geometric Modelling reverse engineering of shapes. While this topic has Laboratory at SZTAKI are reviewed within the been an important topic in computer vision, due to framework of the whole reverse engineering process. the advances in laser scanning, now it has become a

In conventional computer aided design engineering real parts are transformed into particularly important - such as in the the computer representation of objects are such computer models which are suitable automobile industry - real-scale wood or created by means of commands typically to exploit all advantages of modern clay models are often needed, because using an interactive design program with CAD/CAM technologies and transfer real stylists prefer to evaluate real 3D objects 3D graphics. This representation is used objects into virtual reality environments. rather than 2D screen images at reduced for further design, analysis and optionally scale. Another important area of for numerically controlled manufacturing. Typical applications include reproducing application is to generate custom fits to While conventional engineering and redesigning parts, when no original human surfaces, for mating parts such as transforms engineering concepts and drawings or documentation are available. helmets, or prostheses. models into real parts, in reverse In areas where aesthetic design is

ERCIM News No. 44, January 2001 19 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

The Geometric Modelling For conventional mechanical Laboratory at SZTAKI, Budapest engineering objects a different has been conducting reverse approach - ‘direct segmentation’ engineering research since 1995. has been developed. In this context, This activity has been supported by conventional means bounded by various grants including an EU simple analytic surfaces, swept COPERNICUS grant (RECCAD) surfaces and blends. These objects and several smaller ones by the have many faces, which typically Hungarian Scientific Research intersect each other, but the surface Fund (OTKA). Research has been elements are much simpler from focused to develop fundamental geometric point of view. Linear techniques to process and create extrusions and surfaces of accurate geometric models and also revolutions are particularly to exploit the theoretical results in important here, and need to be various applications. (See the web detected and accurately pages of GML and its spin-off reconstructed based on their company Cadmus Ltd. at ‘regularity ‘. Having this limited, http://www.cadmus.hu.) The but very important class of surface process of reverse engineering is elements, it is possible to perform decomposed into the following Figure 1: Car body panel - surfaces reconstructed. segmentation without user steps: data capture, preprocessing interaction. This is illustrated in measured data, merging multiple point This problem is even more difficult for Figure 2, where different colors has been clouds, creating the topology of the object free-form surfaces, where there are assigned to segmented point sets. In by forming a triangular mesh, simplifying unknown smooth edges between the Figure 3, the corresponding reconstructed meshes, segmenting point clouds, fitting individual surface elements, which need CAD model is shown. Note, that surfaces, building boundary to be separated. A new paradigm called smoothness between neighboring surface representation models, adding blending ‘functional decomposition’ helps in these elements is assured, once their likely surfaces and finally ‘beautifying’ models. situations. A free-form object is assumed smoothness has been detected. For Maybe the most crucial part is the to be bounded by relatively large primary example, rotational symmetry was segmentation of the point cloud: subsets surfaces plus connecting surface elements detected for the left part of the object, of points need to be separated from each called feature surfaces, such as blends or consequently the reconstructed torus and other according to their fundamental steps. In the reconstruction procedure first cylinder pieces will be joined smoothly geometric or functional properties. Once we create primary surfaces over in strict mathematical sense. Engineering we have this information, we can fit topologically irregular regions with constraints, such as parallelism, surfaces to the individual point sets, but several holes, and enforce smoothness to perpendicularity, concentricity and others without this or with incorrect assure fair shapes and natural extensions can also be recognized thus the final segmentation the final result is likely to outside the given point regions. Then we model will satisfy these as well. be wrong. For example, if we have a fit feature surfaces, which are constrained subset of points with rotational symmetry, to be smoothly connected to the primary The diversity of reverse engineering but we fail to separate all corresponding surfaces. In this way, based on partial problems has provided and still provides data points and include wrong points, the point information, we can get complete many interesting research topics for the ‘best fit’ rotational surface may be CAD models having the likely original Geometric Modelling Laboratory. By distorted. If we detect only a shrunk surface components by means of which now this technology has reached an subset of the rotational portion, our the object might have been originally advanced level, and it is being directly estimate will not be too accurate. constructed (see Figure 1). exploited in various CAD environments. Our current partners are from the automotive and medical industries.

Links: Geometric Modelling Laboratory at SZTAKI: http://www.sztaki.hu/gml/

Please contact: Tamás Várady Ð SZTAKI Tel: + 36 1 386 8782 E-mail: [email protected]

Figure 2: Benchmark part - point cloud Figure 3: Benchmark part - reconstructed segmented. CAD model.

20 ERCIM News No. 44, January 2001 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

SOLEIL Ð Application of the Radiosity Method to the Physiological Simulation of Plant Growth by François Sillion and Cyril Soler

The goal of the SOLEIL project is to simulate the models is performed by a new technique based on a ecophysiological processes in plants and to combination of hierarchical radiosity and interactively compute their growth and productivity instanciation. This combination allows to perform depending on external conditions, especially the accurate simulations in very large models, which was presence of light. The lighting simulation in the plant not possible before.

SOLEIL is a cooperative research project between INRIA, LIAMA - the French- Chinese Laboratory of Computer Graphics and Applied Mathematics, and CIRAD - the International Research Center for Agronomy and Development. The work carried out in the project can have major impact on at least two different fields: as it is simulates the physiological behaviour of plants, based on calibrated models of growth, it will permit to study the influence of environmental conditions on plant growth for various agronomic applications. it is also of interest for virtual reality systems, because the very realistic plant models it creates can be incorporated into any virtual reality systems such as architecture design programs or video games.

The physiological simulation of plant This virtual plant illustrates the action of lighting conditions onto the shape and size of the simulated models. The non uniform allocation of vegetal matter into the plant model is an growth consists in growing virtual models indirect consequence of the non uniform repartition of light energy in the plant during its of plants from the biological laws that rule growth. As a result the simulated plant reacts as would a real plant: it tends to grow in the plant internal functionality. A direction of the light source and has denser foliage in regions that receive more light. physiological simulator thus accounts for hydraulics inside the plant structure, Instanciation consists in replacing parts Building realistic models of plants is a mechanics that is responsible for the of the plant geometry that share common difficult task since traditional digitization shape of the plant, and the amount of light geometric or radiometric properties by methods (3D-scanners, 3D sampling) and received by the leaves. While the references to shared objects during the modelling tools (CAGD or CAO) can not photosynthesis activity of each leaf indeed simulation. Thanks to the high degree of be used to produce valuable plant models strongly depends on the local self similarity in common vegetation even having a small number of leaves. illumination, all previous plant growth models, this results in important savings The only reliable methods are to grow models did not compute an accurate in memory consumption. Using this tool, plants using either morphogenetic estimation of this quantity because of the we can simulate the effects of light on approaches (L-system, Fractals, ...) or extensive computation cost traditionally plant growth like self-shadowing or physiologically based approaches. Our associated with lighting simulations that heliotropism and thus create models of plant growing approach belongs to the accounts for diffusion and shadowing plants that can react to their environment. second category and thus produces inside the foliage. We have derived a new One other important impact of our work realistic models of plants, that naturally algorithm capable of performing efficient is the possibility to study ecological react to their environment as real plants yet accurate lighting simulations in plants factors on entire plantations, for example do. Such realistic plant models are very consisting of millions of polygons. in order to optimize the distribution of important in the judgment of realism in plants in a field so that they do not shadow synthetic images. They can be used in The method is based on the concept of each other. The applications of the virtual reality systems as well as in instanciation, adapted to the traditional research project are important and architectural modelers or landscape hierarchical radiosity methods for numerous in at least two domains: image design software, and even video games. computing . synthesis and agronomy.

ERCIM News No. 44, January 2001 21 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

Since we not only simulate the shape of global models of canopy reflectance and especially concerning lighting simulation, plants but their complete biological in many remote sensing applications. we will soon be able to bridge the gap behaviour, direct applications of our between these plants and a complete simulation model also concern agronomy. Among direct commercial applications, virtual garden. Small scale or local simulation can help we are planning to integrate the Links: understanding the reactivity of plants to environment sensitivity of plants into the SOLEIL projcet: various external conditions such as commercial version of the CIRAD plant http://www-imagis.imag.fr/SOLEIL/ lighting, water supply or even pruning. growth simulator, called AMAP. This AMAP Plant growth simulator: Large scale simulations allow to virtually software is already used by of http://www.cirad.fr/produits/amap/amapen.html explore the different possibilities for various application fields for integrating Please contact: optimizing plantations in regions where realistic plant models into their own François Sillion Ð INRIA resources are scarce, for instance cultures scenes. It also provides plug-ins for Tel: +33 4 7661 5423 in the tropical forests. Finally, virtual plant modelization/animation software such as E-mail: [email protected] models can also be used to replace costly SoftImageTM and MayaTM. Thanks to the or unpracticable experiments on real plant results we obtained on simulating these Cyril Soler Ð INRIA Tel: +33 4 7661 5429 canopies, which is the case for deriving different kind of interactions, and E-mail: [email protected]

Image-Based Rendering for Industrial Applications

by Samuel Boivin and André Gagalowicz

Several techniques have been recently developed for method that can recover the reflectances of all the geometric and photometric reconstruction of surfaces in a scene, including anisotropic ones, using indoor scenes. The MIRAGES team at INRIA- only one single image taken with a standard camera Rocquencourt designed a new analysis/synthesis and a geometric model of the scene.

The first step of the technique developed by MIRAGES is to capture a single of the scene that we want to reconstruct. The second step is to position the 3D model (including the light sources), that can be built manually using the real image, thanks to a classical modeler such as Maya (Alias|Wavefront). Once the parameters of the camera (position, orientation, ...) have been Figure 1: Real image captured with a Figure 2: Synthetic image generated using automatically recovered, our image-based camera. our technique. rendering software may recover a full BRDF (Bi-directional Reflectance Distribution Function) of all the objects in the scene. Starting with a first assumption about the reflectances of these surfaces, we generate a computer graphics image using an advanced photorealistic rendering software, called Phoenix, developed by our team. Phoenix is a global illumination renderer that can produce photorealistic images in a very short time (faster than current radiosity Figure 3: An application example of our Figure 4: An application example of our software). Phoenix takes advantage from technique in augmented reality: the technique in augmented reality: some the internal architecture of the computer, observer’s position has been modified. synthetic objects have been added and the to greatly accelerate the process of image light sources positions have been modified. A new light source has been added on the desk generation. to generate different illumination conditions. The legs of the desk have been cut. The first image created by Phoenix using the initial hypothesis about the

22 ERCIM News No. 44, January 2001 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

reflectances of the objects, is directly technique uses several complex presents an example of this application, compared to the real one applying a algorithms to estimate the reflectances of with a modification of the viewpoint difference or a more complex operation. these two objects, because they position, while figure 4 shows the same This error image is then used to correct simultaneously interact on one another. scene containing added synthetic objects the estimated reflectances until the result This is a very complex procedure that can under novel illumination conditions. A synthetic image becomes as close as require several minutes of computation lot of other direct applications are possible from the real one. Our image- time, depending on the complexity of the possible, especially in high rate based rendering technique is iterative BRDF of the surface: the most compression of video sequences (with no because it iteratively corrects the BRDF complicated case that we treat is visible loss). Our technique can also be of the 3D surfaces, and it is hierarchical anisotropic surfaces on which highly applied to automatic positioning of because it proceeds with more and more textured objects are reflected. mirrors and light sources, advanced complex assumptions about the reflectance analysis of surfaces, etc. photometry of the objects inside the scene. We show some results in this If a particular reflectance property has presentation, including a real image As a conclusion, we have presented a new been tried for a surface, and if it produces (figure 1) and the full (photometrically image-based rendering technique that is an image that still has important errors on and geometrically) reconstructed very fast and that recovers the BRDF (for all the surface area, this reflectance is synthetic image (figure 2). The mirrors diffuse, specular, isotropic, anisotropic automatically forced to be a more and the aluminium sheet are simulated and textured objects) of all surfaces inside complex one (perfect specular or isotropic respectively as real mirrors and real a scene, including very complex ones. for example). Using these new complex BRDF surfaces: they are not Our method is very efficient and needs assumptions, it is now possible to create simulated as textured objects which is the very few data: one single image taken a new synthetic image that will be again case of most of the techniques developed with a camera and a 3D geometric model compared to the real image, to assess if today. In other words, with one single (including light sources) of the scene are the error between these two images image we know how to reconstruct sufficient for the full reflectances decreased. If the error for an object is mirrors, isotropic and anisotropic recovery. Some future directions are smaller than a user-defined threshold, then surfaces. This procedure possesses the currently investigated in order to develop it is confirmed that its photometry has ability, on one hand, to analyse through new applications and to take into account been correctly simulated and it does not the mirror some objects that were not scenes containing more complex BRDF need to have a more complex BRDF. directly seen in the real image, and on the and objects that combine both anisotropic Moreover, when there are high optical other hand, allows several industrial reflections and texture properties. and thermal interactions between two applications. The first application is called Please contact: objects (suppose for example that a book Augmented Reality and it is very used for Samuel Boivin Ð INRIA is lying on an aluminium sheet and that special in movies in cinema and Tel: +33 1 3963 5186 this book is reflecting on this sheet), our postproduction applications. Figure 3 E-mail: [email protected]

Colour Texture Modelling by Michal Haindl and Vojtûch Havlí ek

The Textures Modelling Project carried out at CRCIM aims at developing mathematical models capable to - UTIA and supported by the Czech Grant Agency, simulate natural colour textures.

Virtual models require object surfaces given natural texture sample (eg, fractals, Modelling multi-spectral images requires covered with realistic nature-like colour random mosaic models, syntactic models, three-dimensional models. However if we textures to enhance realism in virtual SARMA, etc.) the Markov random fields are willing to sacrifice some spectral scenes. Although such models can work (MRF) are the most powerful and flexible information a 3D model can be also with digitised natural textures, ones, because choosing different approximated with a set of simpler 2D synthetic textures are far more convenient, neighbour sets and different types of models without compromising its visual not only because of their huge data probability distribution for the random realism. Random field based models quite compression ratio (dozens model variables a variety of distinct image types successfully represent high frequencies parameters irrespective of the required (sharp edges, patchy images, etc.) can be present in natural textures though low texture size) but can be designed to have generated. Unfortunately MRF models frequencies are much more difficult for certain desirable properties (eg, they can require in general time-consuming them. One possibility how to overcome be made smoothly periodic). Markov chain Monte Carlo synthesis this drawback is to use a multi-scale methods and MRF analysis is complicated random field model. The resolution Among several possible modelling task even for simple MRFs. hierarchy provides then a transition approaches which are capable to learn a between pixel-level features and global

ERCIM News No. 44, January 2001 23 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

resolution factors from either of the models is transformed into the resulting synthetic texture.

Some our synthetic textures reproduce given natural texture so well that both textures are visually indiscernible. The original colour tones are reproduced realistically in spite of the restricted spectral modelling power of the model. The multi-scale approach is more robust and allows far better results than the single-scale one if the synthesis model is inadequate (lower order model, non- stationary texture, etc.). The GMRF multi-scale model seems to be superior to the causal WMRF model for some textures, however the WMRF model synthesis is much faster than the GMRF model synthesis. Another advantage of the WMRF model is its consistent and efficient analytical estimators for all Natural textures from the MIT VisTex (odd rows) and their synthetic counterparts. model parameters which is not the case for the GMRF model. Both proposed features and hence enables to model a model can be found analytically using methods allow large compression ratio large variety of possible textures. Bayesian estimators (the WMRF for transmission or storing texture can be evaluated also recursively if information while they have still moderate Two fast multi-spectral multi-scale MRF required) and the model is easily computation complexity (see our Java texture models developed in the project synthesized from the model equation. Fast applet demo) and avoid any time- synthesize single mono-spectral single- GMRF parameter estimation requires an consuming numerical optimization. resolution factors of a given natural colour approximate maximum likelihood Links: texture using either Gaussian MRF estimator and similarly fast estimators texture demos: (GMRF) or causal wide-sense Markov exist only for a sub-optimal contextual http://www.utia.cas.cz/user_data/haindl/ (WMRF) sub-models. Hence a complex support set. However, also the Fourier virtuous/dema/demtexsyn.html 3D MRF texture model is approximated transformation based GMRF synthesis with a set of simpler 2D MRF models. avoids the time consuming Markov chain Please contact: Michal Haindl Ð CRCIM (UTIA) WSM model parameters as well as the Monte Carlo iterations. Finally the set of Tel: +420 2 6605 2350 optimal contextual support set of the single synthetic mono-spectral single- E-mail: [email protected]

Comparing Real & Synthetic Scenes using Human Judgements of Lightness

by Ann McNamara

Increased application of computer graphics in areas quantify the level of realism achieved by the image which demand high levels of realism has made it synthesis method, and involves running sets of necessary to examine the manner in which images experiments that compare a simple physical test scene are evaluated and validated. Research carried out at to a number of graphical images which are intended Trinity College Dublin explores the need for including to represent it. the human observer in any process that attempts to

The aim of realistic image synthesis is the indistinguishable from an actual scene. of an image synthesis method is evaluated creation of accurate, high quality imagery Reliable image quality assessments are using numerical techniques that attempt that faithfully represents a physical necessary to help determine how close to quantify fidelity using image-to-image environment, the ultimate goal being to image synthesis systems come to comparisons (often comparisons are made create images that are perceptually achieving that goal. Typically the quality with a photograph of the scene that the

24 ERCIM News No. 44, January 2001 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

Figure 1: A graphical rendition of the scene Figure 2: The experimental set-up. used. image is intended to depict). However, The Experiment The main focus and motivation behind this may not reveal specific errors in the Each participant was presented with a this project was the development of a portrayal of lighting and shadows; series of images, in a random order, one process that led to the evaluation of furthermore it could be argued that one of which was the real environment. The computer graphics images with respect needs a measure of fidelity which images presented were the real scene, the to the real world scene they depict. This involves assessment of performance on a photograph and the 9 rendered images. has resulted in a new experimental specific visual task, performance on that The task - that of matching materials in framework that facilitates human task in a real scene can be compared with the scene against a display of originals ˆ comparison of real and rendered scenes. performance on that task in a graphical was chosen because the task has a number scene to give a measure of fidelity. By of attractive features. First, it has been The ultimate goal of this work, which is conducting such a set of experiments, shown that the perception of lightness (the continuing in the Image Synthesis Group based on psychophysics, it can be perceptual correlate of reflectance) is at Trinity College Dublin, is to develop estimated how close a rendered image is strongly dependent on the human visual a computational algorithm to judge the to its original counterpart. system’s rendition of both illumination quality of computer images that mimics and 3-D geometry. These are key features the quality judgement of a human being. The Experimental Set-Up of perception of any scene and are in In effect this would mean designing a The test environment was a simple five- themselves complex attributes. However, form of „visible differences predictor‰ sided cube. Several small objects were the simple matching procedure used here which matches human judgement of an placed within the box for examination. depends critically on the correct image to its real world scene. This would The test scene was then modelled and representation of the above parameters. eliminate the need for further, time rendered on computer. In total ten Therefore, the task should be sensitive to consuming psychophysical studies, by graphical images were considered for any mismatch between the original and incorporating the actual response of the comparison to the real scene, these the rendered scene. HVS into a difference metric. This new included rendered images of varying metric would then give the same results degrees of quality and a digital Results and Conclusions as image comparison by human photograph. A small, front-silvered, high The results show there are indeed observers, facilitating subjective quality mirror was incorporated into the differences between the judgements of comparison of computer graphics set up to allow the viewing conditions to lightness in the real scene and the imagery. facilitate alternation between the two synthetic images. The main conclusions Links: settings, viewing of the original scene or from the analysis of the experimental http://www.cs.tcd.ie/ viewing of the modelled scene on the results are that there is indeed hard . When the optical evidence that the certain images are Please contact: mirror was in position, subjects viewed perceptually degraded compared to the Ann McNamara Ð Trinity College Dublin the original scene. In the absence of the real scene. However, there is no evidence Tel: +353 1 608 3681 E-mail: [email protected] optical mirror the computer representation that the others images in this study are http://www.cs.tcd.ie/people/Ann.McNamara of the original scene was viewed, see perceptually inferior to the real scene. Figure 2. From this we can conclude that the certain images are of the same perceptual quality as the real scene.

ERCIM News No. 44, January 2001 25 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

Facial Repertoires for Animation on the Web

by Zsófia Ruttkay

The CharToon system is a tool dedicated to design editing facilities, the user can define expressions on and animate 21/2D non-photorealistic, cartoon faces a high level and design the dynamism and behaviour for web and other applications. Novel features are of a face. CharToon was developed at CWI in a 3-year based on the concept of a facial repertoire, containing project and will be sold in 2001 by EPICTOID, a CWI ready-to use, but adaptable facial features as well as spinoff company. facial expressions. With constraint-based animation

Humanoids with expressive faces have become popular in social user interfaces as well as virtual reality, telecommuni- cation, Web and entertainment applications. There is demand for tools with which a non-professional user can make a variety of expressive and appealing faces with little effort and resources. CharToon is a vector-graphics based facial animation tool written in Java, with which one can construct faces which can be animated and compose for faces. A player module can generate the animation on the fly, real-time, based on scripted description of the face and the animation. The basic CharToon system was already introduced earlier in ERCIM News No. 35. Figure 1: Three faces made from facial Figure 2: Visemes for a mouth, In the last year, CharToon has been repertoire elements, showing identical and the control points. extended with facilities enabling the re- expres- sions in each row. use of higher-level building blocks, both when making faces and when animating A specific set of facial expressions are Another extension to CharToon is the them. CharToon is equipped with two the visemes, containing mouth shapes for Emotion Disc: a device which allows to ready-to-use, extensible and adaptable talking heads (see Figure 2). By using generate and explore a continuum of sets of repertoires: different viseme repertoires, the lipsync facial expressions, assuming that initially • The facial feature repertoire consists of for different languages or for different the snapshots of the 6 basic expressions a choice for each facial feature such as user groups (eg hearing impaired) can be - joy, surprise, fear, sadness, anger and mouth, eye, hair and facial outline. The automated. disgust - were designed. The further variants may differ in look, in potential elements of the emotion space are motion characteristics or viewpoint. The user starts creating a face by selecting building blocks. He has the freedom to adapt the look and shape of the included features, in order to create an original, new face. • The facial expression repertoire contains still and dynamical expressions, stored as animations.

Due to consistent labeling of the so-called control points which define deformations, the facial and the expression repertoire are orthogonal: an expression can be ‘applied’ for any face made of (adapted) Figure 3: Emotion Disc with the 6 basic expressions. Expressions at * marks are weaker facial feature repertoire elements (see versions of the expression on the perimeter. The expression at location X is computed as bi-linear interpolation of the expressions given at points 1, 2, 3 and 4. The corresponding Figure 1). expression is shown on a well-known ERCIM face.

26 ERCIM News No. 44, January 2001 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

generated by blending (animation Currently we are implementing a new Though CharToon was initially developed parameters of) two of the given emotions version of CharToon, extended with as a research tool, it raised much interest in a certain way. The design of Emotion constraint-handling mechanisms. The in a broad circle of potential users. To Disc is based on an almost 50-year old constraints are to be used in two roles: turn it into a commercial product, the (though debated) observation by • to define characteristics of generic or spinoff company EPICTOID was created Schlosberg, stating that the six basic specific expressions (declarative role); recently, starting to sell CharToon in emotional expressions are perceptually • to aid the user during editing to produce 2001. related in such a way that they can be an animation which fulfils initially set Links: arranged in a two-dimensional space requirements (imperative role). http://www.cwi.nl/CharToon along a circle. The continuous space of Emotion Disc is, accordingly, a round disc By using constraints as conceptual means Please contact: showing a neutral face in the centre and to declare expressions, we will be able to Zsófia Ruttkay Ð CWI maximal expressions on the perimeter address the still open issues of Tel: 31 20 592 4144 E-mail: [email protected] (see Figure 3). The Emotion Disc can be superimposing and concatenating pieces used in all stages of the animators’ work. of animations and experiment with Paul ten Hagen Ð EPICTOID It has also been used as a direct controller different strategies. E-mail: [email protected] of the expression of an avatar’s face.

3D Medical Visualization: Breaking the Limits of Diagnostics and Treatment by Andreas H. König and Eduard Gröller

Three years ago a research team at the Vienna software tool changes the way radiologists and other University of Technology joined forces with a medical specialists diagnose and treat health development division at Tiani Medgraph (one of problems of patients. Two examples out of the variety Europe’s leading companies in Picture Archiving and of new techniques (Virtual Endoscopy and the Communication Systems) in order to develop a specification of transfer functions for Volume medical 3D . Today this comprehensive Rendering) are discussed in this article.

For purposes of medical diagnosis it is founded as well as a highly organs and anatomical cavities using most desirable to have available as much sophisticated understanding of human volume data. Using a real endoscope is detailed information on the anatomical anatomy. To create a ‘complete’ mental always painful or at least inconvenient and pathological features of the patient image of the patients structures, the for the patient. Virtual endoscopy is used as possible. In clinical practice radiologist has to take all available slices as a non-invasive diagnostic technique. examinations by the various scanning into account. It is obvious, that looking It may also be applied in surgical modalities (like Computed Tomography at hundreds of slices (and soon planning, training and exploration of (CT) or Magnetic Resonance Imaging thousands!) is way to time consuming for structures not reachable with a real (MRI)) try to find the optimal trade-off a single patient. Although these new endoscope. All these fields have similar between a high scanning resolution and modalities provide for the very first time requirements concerning the visualization the radiation harm or inconvenience for ever reasonably enough information to system : accuracy, intuitive interaction, the patient. With the recent rise of find even small (and still curable!) fast visualization, and a short modalities featuring high spatial scanning pathological structures, it has become preprocessing time. resolutions (like Multi-Detector CT), impossible to do the diagnosis the special strategies for the investigation of traditional way. 3D Medical visualization The common approach for the huge data sets are more necessary than can help to overcome this problem by visualization of a virtual endoscopy is ever. providing the user with a 3D surface rendering, yielding images close representation of the patients anatomy to a real endoscopy. If external structures Traditionally, a radiologist supposed to reconstructed from the set of image slices. are of interest, diagnose such data sets would look at the Two examples for the application of techniques have to be used. These scanned images slice by slice. By medical visualization are discussed in the methods do not display the exact shape mentally reconstructing the sampled following sections. of the inner lumen very well. For certain information into a three-dimensional applications, eg operation planning of a representation, he would judge on the Virtual Endoscopy transbronchial biopsy, both, the shape of health of the patient. This procedure Virtual endoscopy is an application which the inner lumen as well as outer structures requires the radiologist to have well- deals with the exploration of hollow like blood vessels and the tumor have to

ERCIM News No. 44, January 2001 27 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

Figure 1: Real endoscopy compared with surface shading and volume rendering (left to right).

Figure 3: Rendering of a human hand. A complex transfer function was used to separate bones, blood vessels, ligaments and skin.

Figure 2: Virtual Endoscopy of the aorta and the main bifurcation. Surface Rendering (top left), Volume Rendering (top right), reference slice (bottom left), and endo-luminal view (bottom right). Figure 4: Another transfer function was used to display bone, blood vessels, and skin independently.

be delineated. A new method has been of transfer functions has been developed. Other applications include cardiovascular developed, that allows a quick and easy The specification is usually a difficult task imaging, transplantation-surgery hybrid visualization using overlays of as mapping information for a number of planning, orthopedic diagnosis and different visualization methods like different domains (data range, color, surgery planning, neuro-radiological different surfaces or volume renderings opacity, etc.) has to be defined. In the new applications, vascular diagnosis and with different transfer functions in real approach, the definition of the mapping treatment planning, and a lot of other time on a low-end PC. To achieve real information can be realized independently specialized techniques. One of the future time frame rates, image based rendering for each property domain. A set of goals of computer aided diagnosis using techniques have been used. Virtual specification tools is provided for each 3D visualization is the establishment of Endoscopy images are shown in figures domain, enabling users with different screening examinations. If technology is 1 and 2. levels of experience or demanding time evolved enough, every person in our restrictions to choose an appropriate society will spend a couple of minutes per Advanced Transfer approach for their needs. Real-time year in a scanning modality. The acquired Function Specification during the manipulation of data will be analyzed automatically by an When employing direct-volume- parameters has been proven to be crucial computer system. Potential hazards to the rendering for the visualization of to the specification. patients health will be reported to medical volumetric data sets, typically a transfer experts, who will be able to detect and function is used to classify the data. Data An interactive direct-volume-rendering cure most diseases in time. values are mapped into optical properties display is realized by utilizing dedicated Links: like color or transparency information. . Figures 3 and 4 http://www.vismed.at Especially medical doctors with little show examples of direct volume experience in computers or the rendering displays utilizing transfer Please contact: mathematical background of volume functions specified with this technique. Andreas H. König and Eduard Gröller Ð rendering are usually not able to handle Vienna University of Technology Conclusion Tel: +43 2236 3077 0 complex paradigms for specifying the E-mail: [email protected] high-dimensional functions needed for Just two examples from a variety of the visualization. Therefore a new user- approaches included in the developed interface paradigm for the specification software system were presented here.

28 ERCIM News No. 44, January 2001 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

Real-time Visualization through the Internet by Xavier Cavin and François Cuny

VSP Technology (Visualization Service ProviderTM)is rendering software developed at INRIA Lorraine within a new start-up company born from a technology the ISA research team. The innovative objective of transfer from INRIA and providing services for a VSP is to offer a new approach of real-time realistic and interactive visualization of complex 3D visualization through the Internet as an Application models. The core technology relies on an advanced Service Provider (ASP).

Virtual prototyping is expected by the industries such as automobile, vocabulary, our loading module generates industrial world to offer an important aeronautics, energy, architecture, etc a compact mathematical representation competitive advantage in comparison through an advanced system of geometric of the initial geometry that preserves the with traditional prototyping techniques. optimization, numerical physically based topological properties of the objects. In On one hand, it allows to greatly reduce simulation and real-time images addition, physical and spectral properties the design times and costs of new generation, directly following from the of surfaces and light sources are taken products; on the other hand it may most recent research in the field of into account into the model using standard advantageously assist the marketing and computer graphics. industrial data formats, in order to ensure the high-end management in their making the physical accuracy of its numerical decisions. However, existing visualization The process of Visualization Service representation. solutions continue to suffer from too Providing can be decomposed into three many limitations that have restricted their phases: geometric and physical Then, the propagation of light in the spread and use: physical inaccuracy, very modelling, light transfer simulation and model is computed from a global physical rough , poor interactivity interactive visualization. simulation based on a high performing and over all an incapacity of coping with radiosity algorithm. This algorithm the scale required by industrial For immediate integration purpose, the efficiently performs the calculations using applications. geometric 3D model can directly come high order wavelet functions on from the traditional CAD/CAM/CAE parametric surfaces, and can take VSP Technology provides solutions for modelling tools used in the industry, such advantage of existing multi-processor: it high quality virtual prototyping of very as CATIATM or MayaTM. Unlike standard can thus be applied with extreme large sized numerical models for CAD polygon-based interchange precision and speed to models of very

Interactive design and real-time visualization of an oil and gas control room for ALSTOM: Original model in Maya (top left), Maya visualization (bottom left), VSP visualization (right).

ERCIM News No. 44, January 2001 29 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

large size. By contrast to classical Monte complete visualization solution in an particular, INRIA is involved with France Carlo ray tracing approaches giving static Application Service Provider type of Telecom in the RNRT/VTHD project, images, the radiosity method results in a process. This concept must allow users whose objective is to settle a very high view-independent solution, meaning that to make the most of all the technological network (2.5 Gbits/s) between views can be displayed without requiring power of VSP Technology via the the five INRIA institutes (Paris, Grenoble, any supplementary lighting computation. Internet, and to focus all their efforts on Nancy, Nice and Rennes): this network their know-how by leaving apart all will help us to conduct experimentations The final step consists in interactively hardware and software aspects. No matter of the ASP aspects and to validate this visualizing the illuminated model. Thanks where they are located in the world and approach. to advanced techniques - including regardless of which resources (high-end Links: lighting textures, level of details and or low-end) they are in dispose of, they http://www.vsp-tech.com optimal mesh simplifications - intensively will be able to interactively visualize and http://www.loria.fr/isa using graphics hardware acceleration, modify their numerical models, even the realistic images can be displayed with a most complex and largest ones, with a Please contact: high resolution at a high frequency. never achieved realism. François Cuny Ð INRIA Tel: +33 3 83 59 20 30 Supplementary complex optical effects E-mail: [email protected] (that were indirectly taken into account The objective of VSP Technology is to http://www.loria.fr/~cuny in the original simulation) can be added bring the academic results of the research in real-time, and some modifications can made at INRIA to a competitive Xavier Cavin Ð INRIA be applied to the model, updating the technology. This work will be done in Tel: +33 3 83 59 20 93 E-mail: [email protected] solution accordingly. collaboration with the ISA (Nancy) and http://www.loria.fr/~cavin iMAGIS (Grenoble) INRIA research Furthermore, a strong concept of VSP teams on one hand, and our strategic Technology is to put forward the partner SGI on the other hand. In

The Pond: Trawling the Information Sea

by Lennart E. Fahlén, Jan Humble, Olov Ståhl, Jonas Söderberg and Anders Wallberg

The Pond is a desk-projection based system for approach for querying, presenting and manipulating search, visualization and transactions of data data elements differs considerably from more elements stored in local databases or on the Internet. conventional information systems. Based on a 3D ecosystem metaphor, the Pond

Conventional data search engines (eg results as flocks of aquatic entities inside stored in a database or on an Internet site. Internet search tools such as AltaVista , a 3D virtual pond. The key element in the A number of tag readers have been Yahoo and products such as Oracle interface is the information carrying incorporated into the Pond tabletop, and DBMS) mostly provide textual entities’ ability to make their presence users need only place keyword tags (eg representations of results, and are thereby known and adapt their behaviour based The Beatles or ‘techno’) on one of these somewhat limited to 2-dimensional on how they are handled, such as to initiate a search. Once the search results layout. 3D data visualization systems such acquiring or relinquishing interest towards have been retrieved, a flock of pond as VR-VIBE from Nottingham University user presence. creatures representing the query result is and Q-PIT developed at Lancaster created and inserted into the virtual pond University show some promise in The design is that of a desk-based environment, where the flock starts to building a general overview, but require projection system, a table around which move around and respond to user prior knowledge of the system in order to users gather to perform queries and manipulations. interact and especially navigate. interact with the resulting pond flocks. The back-projected tabletop surface is Entity Flocks The ICE laboratory at SICS experiments touch-sensitive, allowing for direct Each member of the flock has an with new approaches to information interaction with the pond environment autonomous behaviour, controlled by a presentation and manipulation. As part of (see Figure). number of simple navigational rules this effort the Pond has been developed. which, when combined, gives the flock a The Pond is an information system based Queries lifelike behaviour. For instance, entities on an ecosystem metaphor allowing users Users perform queries by using credit card within a specific flock try to stay together to search for data elements stored in local sized tags associated with specific and avoid members of other flocks. This databases or the Internet, and present the keywords that are matched against data allows for easily identifiable result

30 ERCIM News No. 44, January 2001 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

The Pond table with back-projected display in the center. Also visible are a number of tags placed on the table-top.

groupings, even long after the flock was object, moving alongside it and is only Sound created and with the environment vastly visible for a short while. The audio environment of the Pond populated. The flocks also have a consists of both ambient and feedback ‘garbage collection’ behavior that helps Tapping on a creature that is already style sounds. Since the metaphor is that to ensure that the pond will not be selected activates playback of audio of a pond or pool of information, the cluttered with old or uninteresting samples from the corresponding album. soundscape consists of a family of information. As more and more flocks are When this happens the entity will float to aquatic, bubbling, splashing and whirling added, older flocks that haven’t been the surface making it easy to identify the sounds, giving the impression of the data manipulated for some time will record being played. A text will also elements ascending from an abyss of ooze automatically sink towards the bottom of appear inside the pond, stating the record or mud. The ambient sounds fade out the Pond, and eventually be removed. name, artist and title of the track being when a chosen music sample is played, However, users are able to ‘revive’ a played. By repeated tapping on a entity and fades smoothly back in when the sinking flock by interacting with it, in currently being played, users are able to music stops. The interface sounds indicate which case the flock will return to the step through the audio clips available for various actions taken by the users, for surface. the album. Once a track has been played instance placing tags on a reader, dragging the entity will swim away from the entities, etc. Accessing Information surface and return to its flock. Each flock member represents some Conclusion information element that is the result of Building a Selection The Pond is a tool for exploring database a database query. Users are able to access Users can gather entities in creels, specific results in an appealing and inviting this information by manipulating the areas of the pond environment placed , encouraging a collaborative entities in different ways. In the current close to the tag readers. Once inside the space. Based on flocking algorithms to version of the Pond, entities represent creel, an entity is constrained to move group data and direct user manipulation consumer audio items (eg CDs, only within the creel boundaries and a via a touch sensitive display, the Pond MiniDiscs, cassettes etc.), which means new flock is automatically generated, investigates alternative interfaces to that there are two types of information contextually related to the gathered databases, information and transactions available for each entities, the record data selection(s). Placing a recordable-type tag systems. The Pond is an example of an (eg, artist and album title), and the audio on the reader reintroduces the entities Electronic Landscape Artifact developed data itself. Entities in the Pond are easily previously stored on that tag (if any), within the eScape - Esprit long term identifiable because they are texture merging them with the existing selection research project 25377 and the Swedish mapped with the cover image from the in the creel. The link between a Research Institute for Information album they represent. recordable-tag and selected entities within Technology (SITI) Internet3 project ELP. a creel is created when the tag is removed Links: By tapping on an entity (ie touching the from the reader, at which point the creel The Pond project: http://www.sics.se/ice/ display surface), a user selects an entity flock will also be removed from the pond research/projects/pond/index.html and is then able to see information environment. identifying the artist and album title for Please contact: the corresponding record. This Lennart E. Fahlén Ð SICS Tel: +46 8 633 15 39 information is attached to the 3D entity E-mail: [email protected]

ERCIM News No. 44, January 2001 31 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

Advances in Computer-assisted Visualization

by Julian Gallop

Rutherford Appleton Laboratory (RAL) uses under investigation to maintain the quality of this visualization techniques on large amounts of data as service are the use of VR to support 3D visualization, a service to its scientists. Two research topics currently and distributed co-operative visualization.

The main function of the CLRC is to The behaviour of solid materials provide large science facilities that cannot that creep when subjected to be placed in each UK university. RAL loading and high temperatures is hosts a large particle accelerator which a problem which requires the produces muons and neutrons, a investigation of the inside of the synchrotron radiation source to provide material. When welds are X-rays and photons, a laser facility to involved, the problems are more provide pulsed light, and a large NMR complex. Computational facility to provide magnetic waves, all to techniques are becoming more probe the structure of matter. Once reliable at making , experiments using these facilities are and improvements are being designed they produce large amounts of made, but the results need careful data on the structure of samples of examination. Several parameters materials from various specialised are examined including stress, detectors. We also host a large space Figure 1: Visualization of isosurfaces showing the strain and the extent of the science facility that serves earth distribution of micro-organisms in the mouth using creepage at any point. It is observation data from satellites, and act virtual reality. necessary to quickly identify as a data centre for particle physics data critical regions within the from experiments at CERN. These concentrations are high. If one represents materials and examine all relevant facilities produce vast amounts of this by isosurfaces, one has the visual task parameters that may contribute to the disparate data that has to be integrated and of inspecting and interacting with a problem. A static 2D solution to this investigated by teams of scientists collection of sparsely distributed blobs in presents the information, but a moving throughout the UK and Europe. Part of 3D. Particularly when the number of 2D image allows the creepage to be that investigation involves hunting for separate objects is large, VR offers the captured over time. Further, a moving 3D novel or interesting phenomena in ability to quickly perceive the 3D spatial image that the scientist can navigate individual data sets, or more often in relationships that are lost in a 2D through promotes the experience of combining data sets from different presentation - see Figure 1. change and allows scientists to identify detectors or facilities. To identify the novel or interesting in the mass of data available, visualization is a powerful technique, exploiting the human visual system’s ability to search and manipulate objects.

The use of VR to support 3D Visualization Many computer-assisted visualization applications involve the use of 3D and are now sufficiently complex that additional assistance is needed. VR is one technology which promises to help in this since it uses all of the human visual system as it has evolved to detect moving objects in a three dimensional world, rather than just using interpreting static 2D images which is an artificial task at the best of times.

Studying micro-organisms in the mouth can involve investigating the presence and distribution of small regions where Figure 2: 3D VR visualization of creep in a weld using a dataflow visualization package.

32 ERCIM News No. 44, January 2001 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

arbitration. It was therefore possible to design visualization execution networks that handled the application requirements and the collaboration tactics - see Figure 3.

A current project (Visual Beans) takes this further by making use of Java and also component technology. It will build on work at Lancaster University (ADAPT project) which has adopted CORBA to Figure 3: Part of a visualization execution network with a mixture of conventional solve certain problems of delivering visualization modules and collaboration continuous media, where adaptation to modules on the left. varying network and processor conditions is necessary. This development allows critical regions, or more accurately, prototyped. In different projects using this distributed scientists to combine dismiss non- critical regions more quickly approach, 3 different visualization systems components from different visualization - see Figure 2. (AVS, IRIS Explorer, IBM Open DX) systems and different display and have been used with two different VR navigation mechanisms to facilitate the Different system have been systems (PTC & Manchester integration of their knowledge without used for data visualization. In these and University’s Maverik). introducing the barrier of having to learn other examples, a dataflow visualization a specific novel visualization system system and a VR system have been Distributed Co-operative Visualization favoured by another group or discipline. combined. A visualization execution When data from different detectors must Links network is set up for the application in the be combined, groups of scientists need to MANICORAL: http://www.acu.rl.ac.uk/ usual way but with the addition of special combine their complementary specialist projects/manicoral.html modules that communicate 2-way with the knowledge to identify interesting Visual Beans: VR system. The user then interacts with phenomena. A previous project http://www.acu.rl.ac.uk/VisualBeans/ the 3D world through the VR system. The (MANICORAL) developed a prototype ADAPT project: http://www.comp.lancs.ac.uk/ computing/research/mpg/most/adapt.html user has the advantages of VR, such as system that enabled dispersed scientists stereoscopic projection, time-critical to collaborate on such an analysis through Please contact: rendering and 6 DOF input devices, and a common visualization. By basing the Julian Gallop Ð CLRC also the advantages of a dataflow work on a dataflow visualization system, Tel: +44 1235 44 5569 visualization system, where alternative it was possible to develop modules that E-mail: [email protected] visualization solutions may be rapidly handled the necessary dissemination and

Graph Visualization of Complex Information by Ivan Herman

Graph visualization is an emerging branch of structural data representation. The group has information visualization, applicable when data developed the Java-based graph visualization system possess inherent relations, so that they can be Royère, in co-operation with partners at the represented by the nodes of a graph, the edges universities of Bordeaux, Montpellier, and Bath. The representing the relations. A research group at CWI system is freely available for research purposes. focuses on cases where graphs form the fundamental

Graph visualization has many areas of management systems. As the information difficult problems. Even if all elements application. Systems which can be is not always purely hierarchical, more can be displayed, one may obtain an represented as graphs include hierarchies general graphs than trees need to be overall indication but the graph is difficult of files, taxonomies of species, Web site studied. to comprehend (nodes and edges cannot maps, evolutionary trees, and genetic be distinguished). maps. Other applications include class A key issue when visualising graphs in browsers, data structures, state- information visualization is the size of the CWI has developed innovative techniques transition , semantic networks, data. Many applications require analysis to navigate, to filter, and to create PERT diagrams (project management), of graphs with several thousand nodes abstractions from such graphs, in order VR scene graphs, and document and edges. Large graphs pose several to make them usable in practice. A key

ERCIM News No. 44, January 2001 33 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

technique is vital to navigation in large graphs and to identification of important relations among elements. It is particularly important when no ‘natural’ division into subtrees exists. Based on the idea of subdivision, innovative interaction techniques could be developed to enhance the user’s possibilities in analysing the underlying data structures (see Figure 2).

Visual attributes or clusters are generated automatically. If the metric values are Figure 1: Two views of the same graph: without (left) and with (right) emphasis mapping. uniformly distributed, a simple linear mapping to, for example, colour saturation will do. If the distribution is feature is the use of metrics, as a measure reflecting the metric values of the nodes not uniform, the statistical behaviour of associated with a node or an edge. The at its endpoints. In this approach, higher the metric values should be taken into idea is to define the relative importance metric values are considered more account. We have developed techniques of a node or an edge with respect to some interesting and are assigned higher to describe the necessary distributions for semantics. The measure can be saturation values. The result is an the various metrics, and apply them as application specific or based on graph emphasis on the most ‘interesting’ nodes part of the navigational tools. Knowledge structure, for example the degree of a and edges (see Figure 1). Metrics can also of the exact statistical behaviour would node (= the number of adjacent edges), be used to hide elements with low values, save us the process of finding an the size of a subtree, or the flow of thus emphasizing the more important approximation function or building a information in a directed graph. elements in a tree (‘semantic fisheye’). look-up table. Unfortunately, some of these behavioural descriptions are very Such metric values can be translated into Another use of metrics lies in the involved , and require a strong foundation visual attributes like colour and saturation. partitioning of a large data set by in mathematics. Here lies a potential field A technique that we find useful renders subdividing the corresponding graph into of fruitful cooperation between the an edge with continuously shaded colour subgraphs of manageable size. This visualization and the mathematics communities.

All our ideas are incorporated in the graph visualization system Royère, which can be used as the basis for further developments.

Links: CWI’s Information Visualization Project: http://www.cwi.nl/InfoVisu/

Please contact: Ivan Herman Ð CWI Tel: +31 20 592 4163 E-mail: [email protected]

Scott Marshall Ð CWI Tel: +31 20 592 4164 E-mail: [email protected]

Figure 2: The result of simple clustering (skeletoning) combined with metric based colouring, applied to a data set graph with around 600 nodes and 900 edges, which on the screen would look like a uniform blue ‘cloud’.

34 ERCIM News No. 44, January 2001 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

Interactive Visualization of Protein Dynamics by Henk Huitema and Robert van Liere

Understanding the function of a protein is fundamental one or more trajectory files which describe the for gaining insight into many biological processes. coordinates of each individual atom over time. The Proteins are stable mechanical constructs that allow main problem with animating these trajectories is one certain internal motions to enable their biological of temporal scale. Taking large time steps will destroy function. Structural properties of a protein can be the impression of smooth motion, while small time obtained with X-ray crystallography or NMR steps will result in the camouflage of interesting acquisition techniques. Molecular dynamics (MD) motions. simulations at pico/nano-second time scales output

Together with the Department of considered as physically constrained. The can be pegged onto any part of the Microbiology of the University of essential degrees of freedom describe backbone. A measuring stick widget to Amsterdam, CWI has developed an motions which are relevant for the monitor the distance between atoms interactive visualization environment to function of the protein, while the during exploration can be pegged between study protein dynamics. Instead of physically constrained subspace describes two atoms. animating atom trajectory files, essential irrelevant local fluctuations. A number of dynamics is used to filter out non- interactive techniques are added to These techniques have been used to study interesting motions and only visualizing explore the configurational space spanned the dynamics of the Photoactive Yellow correlated atom motions. The basic idea by the essential eigenvectors. One may Protein (PYP). The 3D structure of PYP of essential dynamics is to separate the drag one or more atoms along the is known; it binds a chromophore via a configurational space into two subspaces: eigenvector. The resulting new thio-ester linkage to a unique cystine. an essential subspace containing only a displacement is used to position other When the chromophore receives blue few degrees of freedom, and a remaining atoms in the protein. An angular widget light its conformation changes, which space in which the motions can be to monitor the hinge bending properties affects the overall structure and dynamics of the protein. How to animate internal protein motions?

One hypothesis is that a conformational change on the surface of PYP functions as a signalling device for another protein. Our visualization and interaction techniques help to find which domains in PYP are affected when the chromophore changes state, and which residues are involved in these conformational changes. Animation of the data provides a global overview of the internal protein motions and an indication of concerted fluctuations. Then exploration techniques Figure 1: Line drawing representation are used to gain a deeper understanding of a protein. Arrows indicate the of particular concerted motions. Finally, degrees of freedom of individual exploration combined with measuring bonds. widgets provide quantitative insight into the specific internal protein motions and fluctuations.

In Figure 1 a cartoon representation is used for the backbone (side groups are not shown). The chromophore is shown as a colored and stick representation. Animation of the MD trajectory reveals that the chromophore fluctuates considerably along the most relevant Figure 2: Exploration of the essential Figure 3: Measuring widgets during essential dynamics eigenvectors, and that exploration. eigenspace. these fluctuations are correlated with large

ERCIM News No. 44, January 2001 35 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

motions on the backbone and side chains. conformational freedom of this angle is are glycine-51 and several residues in the In Figure 2, the eigenvector is drawn in responsible for the contraction of the 50- 25-turn. yellow on a selected atom of the helix. The distance widget (shown as a chromophore. The minimum and red-white cylinder) is used to measure the Our work revealed that filtering maximum bounds are displayed on a distance between the 100-loop and the techniques enable the study of internal yellow-white cylinder. The atom is drawn 50-helix. protein motions. Second, active as a semi-transparant sphere. Our exploration of the space spanned by the exploration shows that the motion of the Definite answers to the biochemical selected eigenvectors yields a more chromophore along the eigenvector questions posed cannot be obtained by intuitive understanding than that of induces a contraction of the 50-helix, visualization alone. However, passive viewing of an animation. And causing the N-terminus region to pivot visualization helped to formulate a finally, by exploration in combination around the 25-turn (not shown in the hypothesis which can be tested by further with measuring widgets very specific Figure). Finally, Figure 3 shows an biochemical research. A candidate properties of a motion may be examined. angular widget (drawn as a green domain for the signaling function was Please contact: compass) for measuring the dihedral found to be the N-terminal region. Critical Henk Huitema or Robert van Liere Ð CWI angle that glycine-51 makes with the residues for the conformational changes Tel: +31 20 592 6020 / 4118 backbone. We could see that the large E-mail: {huitema|robertl}@cwi.nl

Physical Deforming Agents for Virtual Neurosurgery

by Michele Marini, Ovidio Salvetti, Sergio Di Bona and Ludovico Lutzemberger

A model of the brain has been developed at IEI-CNR and other types of information gathered from on the basis of data acquired via Computer-assisted anatomical atlases. Tomography (CT), Magnetic Resonance (MR) scans

A model of the brain has been created by continuously in all directions and causes a ‘pressure transmission coefficient’ to means of geometrical and functional data- a resultant that is equal and opposite to each tissue, which simulates the force fusion so that a characterization of each the modifying charge. An approach of this dispersion and is transmitted from one of the internal structures is possible. A set type necessitates some corrections to spring to another. Moreover, the effects of deforming agents has been placed over adapt the discretization to the real-world induced by the presence of a liquid mass, the geometry of the three-dimensional situation. This can be done by assigning which in our study can be considered as model in order to simulate the evolution of different pathological intra-cranial phenomena such as haemorrhages, neoplasm, haematoma etc. and to describe the consequences that are provoked by their removal as well as the influences they have on the anatomical and neuro- functional structures of the brain. This model can be used effectively in virtual neurosurgery. The deforming agent action is an isotropic pressure that can be studied using a number of ‘lines of force’ traced radially, within the geometry of the model. These lines start from a fixed region, chosen as the source of expansion, to the bone boundary limits, as the skull represents a physical limitation to the cerebral expansion phenomena.

The tissue elastic response can be simulated as a series of springs, each of which is associated with the underlying tissue section. The reaction of a tissue to Example of the system interface when simulating an intracranial expansion phenomenon, the deforming action develops applied to a healthy brain CT scan.

36 ERCIM News No. 44, January 2001 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

static, must also be taken into account. regional contours define the knots that are interactive system has been implemented The physiological mechanism adopted used as control points for the application which includes both single data-set slices also favours the use of a spring model in of the geometric deformation. Each and three-dimensional deformation tools the case of those overlapping ‘lines of section comprised between two adjacent to interactively control the virtual actions force’ sections that contain liquid. Each control points lying on the same line is (see Figure 1). straight line coming from the initial point associated with a single spring that rules Please Contact: of development represents a direction of the elastic behaviour of the underlying Ovidio Salvetti Ð IEI-CNR the force field whose vectors are oriented region. The generic spring is then Tel: +39 050 315 3124 from the centre of the source of the characterized by a set of parameters E-mail: [email protected] expansion to the external contour. The whose values depend on the regional intersections where the straight lines cross characteristics. Under AVS/Express’, an

Visualization of Complex Dynamical Systems in Theoretical Physics by Anatoly Fomenko, Stanislav Klimenko and Igor Nikitin

A research team at the GMD Institute for Visualization dynamical systems in theoretical research, like string and Media has developed methods theory or time-symmetric electrodynamics. of computer visualization to understand complex

Investigation of dynamical systems in known of them are the so-called Euler, has topologically different types of theoretical physics is often reduced to Lagrange and Kovalevskaya cases. In motion. analysis of properties of certain spite of the fact that during the last 100 geometrical objects, particularly, surfaces, years a large number of papers and even String theory is a contemporary model of characterizing the physical state of the monographs have been devoted to the elementary particles, which presents them system. Singularities of such surfaces investigation of their properties, since as a system of point quarks, connected by correspond to critical phenomena in only recently the analysis of their global a string-like tube of chromodynamical underlying dynamics. Methods of topology is performed. It turns out that field. The evolution of strings is governed computer visualization can help to computer and visualization methods work by the following dynamical principle: understand a structure of singularities on very effectively in this field, and surfaces, swept in motion of the string characteristic surfaces, shed light on deep sometimes allow to solve new purely through the space-time, called mechanisms of complex physical analytical problems. The project worldsheets of strings, are extremals of processes and give inspiration necessary combines the two approaches: classical area functional in Minkowski space-time. to find a solution of the problem. analytical methods and computer In other words, worldsheets are non- visualization to investigate such Euclidean analogs of soap films. Topological Atlas of Integrable Cases integrable cases. Particular result, in Rigid Body Dynamics obtained in the project, is visualization of One of the interesting properties of such In Hamiltonian mechanics of a rigid body bifurcation for Kovalevskaya surfaces is existence of topologically there is a wide class of special (integrable) top, a surface, subdividing the phase space stable singularities, correspondent to cases, for which the motion of a body has of the system to regions, where the top points of energy concentration inside stable, regular properties. The most well- elementary particles. Recent research,

Figure 1: Bifurcation diagram. Figure 2: Fold on worldsheet. Figure 3: Cayley surface.

ERCIM News No. 44, January 2001 37 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

based on visualization of worldsheets, result, at a certain critical detects solutions of new type, whose value of velocity of density of energy changes sign in passage colliding particles, equal through the singularities. Correspondent to 93.7% of light worldsheets have folds, like shown in the velocity, solution figure. Slices of the fold by planes of undergoes elementary constant time produce the following string catastrophe of evolution: appearance of string bifurcation (splitting): in component with total zero energy from point B one solution (A) vacuum in point A, its recombination with splits into three (C,D,E). main positive energy string component Figure 4: String dynamics. Pictures of scattering for in point R and disappearance of resulting solutions A,D have a zero-energy string component in vacuum symmetry under spatial in point D. In this way the usage of formulated in works of Wheeler and and temporal reflections, while such visualization methods allows to Feynman in 1949, who had shown full pictures for solutions C,E are not demonstrate instability of vacuum state mathematical consistency of such theory. symmetric, but transformed one to in string microworld. Since this time the model of Wheeler and another by reflections. Thus, the Feynman attracts the interest of bifurcation is followed by an effect of Time-symmetric Electrodynamics theoretists. Particularly, structure of spontaneous breaking of mirror Theoretical physics knows several solutions in a problem of 1-dimensional symmetries in studying dynamics. examples of dynamical systems, which relativistic scattering of 2 particles in Links: incorporate direct influence of future to time-symmetric electrodynamics is Homepage of the project: the present, ie the systems with equations defined by properties of stationary http://viswiz.gmd.de/~nikitin/viz.html of motion x’’(t) =F(x(t+1)) or similar. manifold of Poincare map. Visualization One such system, called time-symmetric of this manifold shows, that it has a shape Please contact: electrodynamics, was introduced in the of Cayley surface, one of elementary Igor Nikitin Ð GMD Tel: +49 2241 14 2137 beginning of 20th century, and strictly surfaces in the catastrophe theory. As a E-mail: [email protected]

Simulation and Visualization of Processes in Moving Granular Bed Gas Cleanup Filter

by Pavel Slavík, František Hrdli ka and Ondfiej Kubelka

Computer graphics in last years has been used in graphics is used to visualize results of these many new application areas. Its use allows the user simulations. In the project described further we deal to investigate many phenomena in very detailed ways with a specific application where the use of computer that was not possible before. Many applications are graphics (namely its visualization capabilities) is rather linked-up with simulation processes and computer new.

The project GACR 101/99/0647 process. The coal is processed by means absorption capabilities the granules (Simulation model of filter with vertical of special technology that converts it into remove the polluted part of the gas. In moving layer) is solved in collaboration gas. This gas before it is used in boilers order to remove the granules that have between the Czech Technical University must be purified. One of the most used reached their level of saturation and thus (Department of Fluid Dynamics and purification technologies is based on the are not able to absorb more pollutants the Power Engineering and Department of moving granular bed gas cleanup filters. whole volume of the filter (virtually all Computer Science and Engineering) and granules) are moving downwards to the the Institute of Chemical Technology in These filters consist in principle of a filter lower outlet. The saturated granules are Prague. body (that might have a various shape that continuously removed from the filter and influences the efficiency of the cleaning the new (clean) granules are added The project was targeted to the simulation process) and from the large number of through the top filter inlet. of moving granular bed that has cleanup spherical granules that fill the volume of filter. These filters are used mainly in new the filter body. The gas that has to be Design of these filters is rather combustion technologies that increase the cleaned goes under pressure through this complicated as there are many parameters efficiency of the electricity production volume of granules. Due to their that influence the efficiency of the

38 ERCIM News No. 44, January 2001 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

the user to extend the system capability to investigate the gas flow in 3D (and thus to investigate absorption conditions in any place of the filter). Also areas where stagnation zones can occur are easily detectable. In order to prove validity of the model it was necessary to compare results obtained by means of the model with real results measured on the physical model. The real data were obtained from National Taiwan University where the physical experiments have been performed for quite a long time. The results obtained by means of simulation (and subsequent visualization) were very encouraging. The match was very good see the figure. It is possible to see how group of granules is moving through the filter (and how the initial volume of granules is distributed in the filter volume within certain period of time). Having this Visualization of gas purification. sort of simulation and visualization tool it is possible to design in very quick and cleaning process. The most important appropriate velocity values that lead to effective way the described type of filters. ones are the speed of the granules within optimal performance of the filter. It is possible to experiment with various the filter, shape of the filter and absorption shapes of filters and various speeds of capability of granules. Especially the Traditional approach to the filter design granules in the filter. shape of the filter is of a key importance is to construct a physical model of such as it determines the size and location of a filter and perform set of measurements The system has been implemented in C++ stagnant zones where granules do not on this model. It is obvious that this for MS move with appropriate speed or they do solution is expensive and also time Windows9X/NT platform. not move at all. The existence of the consuming. Solution to this problem is to Please contact: stagnant zones decreases efficiency of the develop a computer model that contains Pavel Slavík Ð Czech Technical cleaning process as the saturated granules both simulation and visualization part. In University are not removed from the filter (either at our project there was developed a model Tel: +420 2 2435 7480 all or very slowly). To achieve some kind based on the use of discrete element E-mail: [email protected] of optimal solution it is necessary to method. The algorithms developed were establish proper filter shape and designed for 3D environment what allows

Watching Chromosomes during Cell Division by Robert van Liere

Visualization is an important tool to provide insight techniques to analyze the condensation and in the structure and functioning of time dependent 3D decondensation of chromosomes during cell division. processes in live cells. In cooperation with the To this end, an interactive VR enabled visualization Swammerdam Institute for Life Sciences in system specially equipped for time dependent 3D data Amsterdam, CWI has applied a number of visualization sets was developed.

In order to understand the biological techniques. Whereas 3D volume data may developed. Next to basic support tools it processes underlying control of gene be analyzed with traditional volume comprises an open ended architecture in expression, the organization of the cell rendering and isosurfaces, the which application specific demands can nucleus must be studied. Time dependent visualization of complex 3D time be easily incorporated, and can be used volume data describing specific processes dependent multi channel data demands a in fully immersive, semi-immersive, and of the living cell are available by now, combination of techniques not found in desktop environments. but their structural analysis in space and standard systems. Therefore a special time is next to impossible with traditional visualization system, called Proteus, was

ERCIM News No. 44, January 2001 39 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

Using 3D confocal microscopy, biologists voxel size is 90x90x700 nanometer. The and rotation of the cell nucleus and by have been able to produce time dependent time step between scans ranges from 30 noise of the data. Hence, the movement 3D data sets of DNA/chromatin in the seconds in the beginning of the process of the cell was filtered out using a center nuclei of living cells. Time series of 3D to 180 seconds at the end. The required of mass estimation, and the noise was images are made of the process of storage size was 98 MB. suppressed using a low pass filter. decondensation of the chromatin after cell Isosurfaces were useful for analysis of the division (mitosis). During mitosis In order to get an overview of the data, motions of dense parts in the chromatin chromatin is densely packed in volume rendering was used because this data. Transparant isosurfaces were chromosomes. After mitosis, part of the presents the data in a natural and intuitive introduced to show the boundary of the chromatin decondensates to form a new way (see Figure 1). This gave insight into nucleus using a low iso-value (see Figure nucleus. The data set consisted of a series the global structure of the data. Visible 2). These dense parts moved apparently of fifty stacks containing thirty optical tracking of the structures in the nucleus mainly outward, without much sections of 256x256 voxels each. The is complicated by the overall movement reorganization inside the nucleus.

Automatic object detection was combined with manual editing. The extensibility of the Proteus system allowed for integration of the feature detection and tracking algorithms into the visualization of the raw data. The tracks found can be manually edited using direct manipulation in places where the algorithm was not capable of calculating the tracks. In many cases the found tracks could be improved by visual inspection. By using the tracking algorithm in combination with the track editing tool the main tracks of dense chromatin spots were found (see Figure 3). Figure 1: Volume visualization of the data Figure 3: Tracks of dense parts in the data during the decondensation process. for 22 time steps. The following lessons were learned from this case study. First, the visualization system used must be flexible enough to cooperate with the other parts of the process. Hence, Proteus was designed as an open system in which the output or other analysis tools can be incorporated. Second, since each visualization technique has its strengths and weaknesses, a whole repertoire of techniques should be offered for use. Finally, for visualization to be an effective tool for collaboration and communication as well, results found in a highly sophisticated environment should be recordable, such that the results can be reproduced for discussion and publication.

Please contact: Robert van Liere Ð CWI Tel: +31 20 592 4118 E-mail: [email protected] http://www.cwi.nl/~robertl/

Figure 2: Isosurfaces in the filtered data. A low threshold value was used to show the boundary of the nucleus (green transparent isosurface). The dense parts of the chromatin are shown in red.

40 ERCIM News No. 44, January 2001 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

The blue-c Project by Markus Gross and Oliver Staadt

The blue-c is a joint research project between several set of video cameras. The project has started in April institutes at ETH Zurich. Our goal is to build a 2000 and its first phase is expected to be completed collaborative, immersive virtual environment which by spring 2003. will eventually integrate real humans captured by a

Mastering the rapidly changing computing and communication resources is an essential key to personal and professional success in a global information society. The main challenge consists not only in accessing data, but rather in extracting relevant information and combining it into new structures. The efficient and collaborative deployment of applications becomes increasingly important the more complex and interactive tools we have at our disposal. Today’s technology enables information exchange and simple communication. However, it often fails in the promising field of computer enhanced collaboration in virtual reality environments. Some improvements were made by coming-of- age virtual reality systems that offer a variety of instrumental tools for stand- alone visual analysis. Nevertheless, the crucial interaction between humans and Figure 1: The blue-c: Design study of the envisioned system depicting projection and life video stream acquisition. virtual objects is mostly neglected. Therefore successful models of truly computer supported collaborative work are still rare. • new interaction metaphors between • three-dimensional human inlays humans and simulated artifacts of reconstructed from video images Real Humans in Virtual Worlds functional and/or behavioral nature. • voice and spatial sound rendering The blue-c project aims at investigating • architectures for a new generation of virtual design, The Technology real-time image processing and modelling and collaboration The blue-c system foresees simultaneous rendering environments. Three-dimensional human acquisition of life video streams and • a flexible communication layer representations will be integrated in real- projection of virtual reality scenes. Color adapting to time into networked virtual environments. representations with depth information of • a scalable hard- and software The use of large screens and cutting edge the users will be generated using real-time architecture for both fixed and mobile projection technology creates the image analysis. The computer generated installations impression of total immersion. Thus, graphics will be projected onto wall-sized • an advanced application programming unprecedented interaction and screens surrounding the user, allowing interface. collaboration techniques among humans him to completely enter the virtual world. and virtual models will become feasible. Multiple blue-c portals, connected by Architectural Design highspeed networks, will allow remotely Today’s architects work with both low- Computing – Communication – located users to meet, communicate and scale physical models and computer Collaboration collaborate in the same virtual space. generated images of buildings. But the The blue-c features are: advantages of full immersion, involving • full immersion of the participants in a The blue-c system will include: visual, acoustic, and haptic senses remain virtual world • a fully immersive three-dimensional widely unexplored. The blue-c system • three-dimensionally rendered human stereo projection theatre will provide the technology for a variety inlays, supporting motion and speech • real-time acquisition of multiple video of new collaborative design, management in real-time streams and refinement procedures between the

ERCIM News No. 44, January 2001 41 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

architect, his customer and third-party common virtual environment. We makes of the blue-c system an essential experts. envision the blue-c technology as the tool for future remote collaborative space of the future. diagnosis. Next Generation Product Development Computer-aided Medicine Blue-c is designed by an interdisciplinary A new generation of development Collaborative diagnosis and telepresence team of researchers, including people environments will emerge in the near open new possibilities in the field of from the several ETH institutes. future. The underlying metaphor is a computer-aided medicine. Remotely Links: virtual space which is adaptable to various located experts can participate in the Project home page: scenarios and to different stages of the diagnostic process. Highly immersive http://blue-c.ethz.ch/ product development process. A network virtual environments allow researchers of globally distributed development and medical students to take interactive Please Contact: spaces enables engineers and product journeys into the human body. The Prof. Markus Gross Ð ETH Zurich Tel: +41 1 632 7114 managers, field technicians and sales unprecedented quality of natural human E-mail: [email protected] consultants to meet and collaborate in one interaction in a distributed environment

Augmenting the Common Working Environment by Virtual Objects

by Wolfgang Broll

While people are used to interact in their natural three- workspace by virtual 3D objects. It provides a basis dimensional environment, they often face difficulties for natural 3D interaction by using real world items to work efficiently using 3D applications. This as tangible interfaces for virtual world objects. Thus especially applies for unskilled or inexperienced intuitive collaboration between multiple participants users. The Virtual Round Table is a new innovative is supported without sustaining existing collaboration environment enhancing the user’s communication skills.

In order to give a large number of people access to 3D technology to improve their daily working environment, interacting in a 3D environment should be as intuitive as interacting within the real world. The Virtual Round Table tries to realise an environment, which provides the basis for intuitive collaboration. It extends the users‚ physical environment by artificial 3D objects. This approach allows us to represent even complex dependencies within the common working environments providing the basis for a collaborative planning and presentation environment. The Virtual Round Table realises this environment using a location independent and inexpensive approach. Simple Virtual Round Table scenario in the area of urban planning.

The Virtual Round Table allows all participants of a work group to stay and collaboration and cooperation space that includes both, physical as well to act in their natural environment. mechanisms within regular meeting as virtual objects. The visualization of an Additionally familiar verbal situations and extends them by arbitrary individually adapted stereoscopic view communication and interaction are not virtual scenarios (see Figure). for each user is realized by the multi-user restricted by the system. Intuitive virtual reality toolkit SmallTool. In order interaction is a central aspect for the user‚s Visualization to provide a seamless integration of the acceptance of the system. Combining real The visualization of a synthetic scene real and virtual world, the virtual view of and virtual objects into a conceptional within a real world working area is each user has to be visualised according unit enables users to intuitively grasp and realised by using see-through projection to his or her current real world location re-arrange virtual objects. Thus the glasses. Thus each user receives an and viewing direction in real-time. This Virtual Round Table emphasizes common individual visual impression of the work requires a continuous tracking of the

42 ERCIM News No. 44, January 2001 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

user‚s head position and orientation. Due position and orientation to the appropriate concerts, fairs, theatre plays and product to the high sensitivity of the human vision properties of the associated virtual object. presentations. Other application areas are system, the device‚s position and The superimposition and synchronization architectural design, urban planning, orientation detection mechanisms have to of the real items with virtual objects leads disaster management (eg coordinating be highly accurate. Moreover, the device to a fusion of the artificial object and its action forces for forest fires) as well as should be insusceptible against external physical reference into a symbiotic unit. cooperative simulation and learning ascendancies. These interaction units allow users to environments. interact on virtual objects directly and Our current prototype is realised on PCs naturally. Conclusions and Future Work with high-end stereo-capable graphics Natural interaction mechanisms are accelerators. User head tracking is This requires an additional mechanism in facilitated by using real items as tangible provided by an InterSense Mark 2 order to recognise and track the position interfaces to virtual objects, without tracking system. However, this tracking and orientation of these real world items. restricting the users‚ common interaction device does not fulfil the requirements of Since we do not want to use designated and communication mechanisms. In our light-weight and non intrusive system. representatives based on markers or future work we will elaborate these Thus a sourceless and wireless six degree attached sensors, we use a computer interaction units and evaluate multi-modal of freedom inertial tracking device based vision based approach. Real world items interaction mechanisms. We will extend on the existing MOVY prototype is are tracked by analysing video images of the inertial tracking device to support six- developed. head-mounted or fixed cameras. Our first degrees-of-freedom and enhance the approach is based on simple pattern and computer-vision based system. Tangible Interfaces luminance detection. Future Links: In our approach we use real world items developments however, will use feature Collaborative Virtual and Augmented as natural tangible interfaces for 3D vectors for object recognition and Environments Group at GMD: interactions. By associating manipulable localisation. http://fit.gmd.de/camelot virtual objects with arbitrary physical items, we create an intuitive interaction Application Areas Please contact: Wolfgang Broll Ð GMD mechanism. This interaction mechanisms We currently evaluate the Virtual Round Tel: +49 2241 14 2715 allows users to interact with virtual 3D Table environment in the area of E-mail: [email protected] objects similar to real world items. It is collaborative planning applications. This realised by directly mapping the items‚ includes the planning of events such as

Levels of Detail in Physically-based Real-time Animation by John Dingliana and Carol O’Sullivan

The demand for realism in interactive virtual animation. An ongoing project at the Image Synthesis environments is increasing rapidly. However, despite Group (ISG) in Trinity College Dublin (TCD) addresses the recent advances in computing power, it is the problem of optimizing this speed-accuracy trade- necessary to make certain simplifications, sacrificing off by using adaptive processes for computationally accuracy in order to meet the demands of real-time intensive tasks such as Collision Detection.

Physically-based modelling concerns addresses this problem by using adaptive, simplification, based no doubt on worst itself with emulating the properties and time-critical algorithms to deal with some case scenarios, is usually excessive as we behaviours of objects in the real-world. of the more computationally-intensive would end up simplifying the entire Objects in a physically based virtual processes in the production of an animation for the sake of what is likely environment interact with each other animation. to be relatively fewer snapshots of high based on rules that are modelled on the complexity. It would be more desirable laws of physics. Unfortunately, given the A naive approach to ensure real-time to be able to dynamically adapt levels of finite nature of machines in comparison frame rates would be to simplify the scene simplification based on the workload at to the infinite complexity of the physical to such a degree that it would always be every stage of the animation. Even better world, we must always accept that no possible to perform the necessary would be if we could strategically matter how detailed our model, it will be calculations involved in producing each simplify, in every frame, certain parts of at best an approximation. This is frame of an animation. However, it is the scene that were less important. particularly true in the case of real-time frequently the case, particularly in animation where the requirement is to interactive animations, that the One consistently time-consuming part of produce frames on the fly at a rate of at complexity of a scene will change physically-based animation is detecting least 20 per second. Current work at TCD significantly over time. Such pre-emptive and handling the case where two or more

ERCIM News No. 44, January 2001 43 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

deals with collisions based on their perceptual importance, allocating more time to the processing of higher priority collisions. Criteria for judging the importance of particular objects are based on paychophysical tests to determine the effects of states and conditions which might affect user-perception of collisions.

Ongoing Work A library for adaptive collision detection Collision Resolution at different levels of Detail. Image shows the results of interrupting the collision handling process at different stages. Red spheres indicate the nodes which will be and response entitled ReACT (Real-time treated as colliding. Green lines indicate the direction in which impulses are finally applied. Adaptive Collision Toolkit) is currently under construction and will allow the low objects in the simulated world come into the object. The volume model is used level procedures for collision handling to contact. It is often through such events instead in calculations involving the be used in more specific applications. that the user is able to interact, directly or objects state and behaviour in the virtual Extensive user-tests have been performed indirectly, with the environment. Thus, it world. Typically, it consists of unions of to determine user perception of collisions, is important that the effect of such events simpler volumes (eg spheres) which are and an eye-tracking device is being used is modelled accurately as it will stored in hierarchical tree structures so to track which collisions a user is looking significantly impact upon the believability that it is possible to localize which part at. Further user tests will be undertaken of the animation from the user’s of colliding objects are touching. The with the specific goal of determining the perspective. On the other hand, such process of collision detection involves perception of collision response. More events are not only quite frequent but also intersection tests using the volume models efficient volume representation schemes require a large amount of processing time at different levels of detail, traversing the using heterogeneous unions of to deal with appropriately. It seems volume tree to get a better approximation subvolumes are also being investigated. apparent that collision handling is one of of the actual contact points. Interrupting All of this goes towards the goal of those processes where significant gains this tree traversal at different stages allows optimising the speed accuracy trade-off can be made by carefully balancing the us to obtain approximations of the contact whilst guaranteeing real-time animation. speed-accuracy trade-off. at different levels of accuracy. Thus, we Links: can not only process individual frames http://isg.cs.tcd.ie/ We employ an approach which uses but also individual objects in a single hierarchical multi-resolution volume frame at different levels of detail. Please contact: models to represent objects in the scene. John Dingliana Ð Trinity College Dublin This is an approximation of an object’s Collisions in a scene are prioritised based Tel: +353 1 608 3679 E-mail: [email protected] volume at different levels of detail and is on their potential effect on the separate from the model used in rendering believability of the animation. A scheduler

Static Solution for Real Time Deformable Objects With Fluid Inside

by Ivan F. Costa and Remis Balaniuk

Long Elements Method (LEM) is a new method for static solution for elastic global deformations of physically based simulation of deformable objects, objects filled with fluid based on the Pascal’s principle suitable for real time animation and virtual and volume conservation. environment interaction. The approach implements a

Physically based simulation of behavior, volumetric modelling what is a good approximation for deformable objects is a key challenge in (homogeneous and non-homogeneous biological tissues. The volumes are Virtual Reality (VR). LEM was materials) and scalability (high accuracy discretised in a set of Long Elements conceived for soft tissue real time when needed). The choices made to (LE), and an equilibrium equation is simulation, particularly for surgical define the method were driven by these defined for each element using bulk simulation. The priorities in this kind of priorities. variables. The set of static equations plus application are: unrestricted multi-modal the Pascal principle and the volume interactiveness, including interactive The approach is based on a static solution conservation are used to define a system topological changes (cutting, suturing, for non-linear elastic deformations of that is solved to find the object removing material, etc), physically based objects filled with uncompressible fluid, deformations and forces.

44 ERCIM News No. 44, January 2001 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

The discretisation adopted by the deformation problem in an elegant method, based on the long elements, and simple way, independently of has two main advantages: the number other simulation aspects as of elements used to fill an object is movements (translation, rotation, one order of magnitude less than in a dynamics integration, etc), topology discretisation based on tetrahedric or changes and collision handling. To cubic elements; the graphic and the obtain a generic simulator the LEM haptic feedback can be directly must be integrated to methods suitable derived from the elements, and no to handle these aspects. To add intermediate geometric representation movements to the simulated objects is needed. The use of static instead of we intend to implement a skeleton- PDE equations avoids all the like structure inside the object and to problems concerning numerical base the dynamics estimation on it. integration, ensuring stability for the Larges deformations, as twisting and simulation. bending will be then possible. Soft tissue (cube) touched by a rigid probe (sphere). No pre-calculations or condensations To facilitate modelling and collision are used, in order to enable real time detection we intend to couple the topology changes. The idea is to fill the surface, or in other words, the variation LEM to implicit or parametric surface volume with long elements, to define of volume due to the elements touched methods. The deformations of the model equilibrium equations for each element by the contact have to be equal to the will be simulated in a set of LE meshes based on the stated principles and to add sum of the volume created by the defining deformable 2D sheets and global constraints in order to obtain a dislocation of all untouched elements rendered as a 3D volume through an global physical behavior. to ensure the volume conservation. implicit of parametric formulation of the object shape. Different meshing strategies can be LEM was used to implement a generic conceived to fill the objects. We use soft tissue VR simulator. The simulator The transformation from 2D sheets to 3D Cartesian meshes. The long elements are was implemented in C++ in a Windows surfaces can be done using texture parallel to the Cartesian axis and cross NT platform. The system main modules mapping techniques. Topology changes the object. One mesh defines one are: are typically task-oriented and use task- deformable face for the object. Multiple • Model definition: the geometry and the specific interfaces. We intend to meshes can be combined to obtain a full physics of the objects to be simulated implement a surgical interface based on 3D deformable object. are defined using available information. medical procedures (cutting, suturing, This information can be derived from removing material, etc). A long element can be compared to a non- segmented real data or from linear spring fixed in one extremity and intermediate models. The volumes are The method can also be useful to validate having the other extremity attached to a meshed to define the long elements. and to adjust estimated biomechanical point in the movable object surface. These • Simulation loop: solves the system parameters in biologic tissues. springs are relaxed when the solid is not equations to obtain the deformed shape Experimental data can be compared to touched (not deformed). The spring of the objects. simulation results. constant depends on its length. A long • Rendering loops: graphic and haptic element does not occupy real space and loops enable the interaction with the The project is carried out among the has no mass. The real space inside the objects. SHARP research groups at INRIA, solid is occupied by some incompressible CATSS-Stanford University-USA and fluid with mass (density). In a standard dual 700MHz PC one Catholic University of Brasilia-Brazil. iteration of the simulation loop takes Links: To re-establish the only one entry solid, about 0.05 seconds for a 600 elements Sharp Research Group at INRIA: two border conditions are applied: mesh. Note that a given discretisation in http://www.inrialpes.fr/sharp • Pascal’s principle says that an external our method gives far more precise CATSS-Stanford University, USA: pressure applied to a fluid confined deformations than the same number of http://www.catss.stanford.edu within a closed container is transmitted tetrahedra or cubes in a standard mesh. Please contact: undiminished throughout the entire The global deformations are physically Ivan F. Costa Ð INRIA fluid consistent and important phenomena such and Catholic University of Brasilia • The fluid is considered incompressible. as the movement of all parts of the solid Tel: +55 61 356 9000 It means that the volume conservation due to the preservation of volume are E-mail: [email protected] must be guaranteed when there is some automatically produced. The method Remis Balaniuk Ð CATSS-Stanford external contact to the object. The presented in this paper is intended for University and University of Brasilia volume dislocated by the contact will being part of a more general VR Tel: +1 650 724 7397 cause the dislocation of the entire simulator. The method solves the E-mail: [email protected]

ERCIM News No. 44, January 2001 45 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

Exploring Virtual Prototypes using Time-Critical Rendering Techniques

by Enrico Gobbetti and Riccardo Scateni

Scientists at CRS4, the Center for Advanced Studies, are represented as multiresolution meshes. In Research and Development in Cagliari, Sardinia, Italy, collaboration with other European partners, this have developed a time-critical rendering algorithm technique has been applied to the visual and that relies upon a scene description in which objects collaborative exploration of large digital mock-ups.

When undertaking a large and lengthy the human perceptual system requires the memory requirements, because of the engineering or architectural project, it is ability to trade rendering quality with need to store multiple level-of-details vital to verify frequently the possible speed. Ideally, this time/quality conflict (LOD), and quality of results, because of consequences of decisions taken during should be handled with adaptive the NP-completeness of the problem. We the design phase. The aim of virtual techniques to cope with a wide range of recently demonstrated that these prototyping research is to allow architects, viewing conditions while avoiding worst- drawbacks can be overcome when using engineers and designers to work on digital case assumptions. The presence of appropriate multiresolution data structures mock-ups which simulate the visual moving parts and the need for interaction (TOM, Totally Ordered Mesh) which appearance and behaviour of objects on between virtual prototyping tools limits make it possible to express predictive a computer. The expected benefits of the amount of pre-computation possible, LOD selection in the framework of virtual prototyping technology include: a implying a run-time solution. continuous convex constrained substantial reduction in development time optimization. and manufacturing costs, thanks to a Time-critical rendering of scenes reduced need for expensive physical composed of many objects is an open Visual Exploration mock-ups; the ability to maintain digital research area. The traditional approach Our time-critical multiresolution scene mock-ups in sync with the design, and towards rendering these scenes in a time- rendering algorithm has been therefore the possibility to use them for critical setting is to pre-compute a small implemented and tested on both UNIX documentation and to help the dialogue number of independent level-of-detail and Windows platforms. The TOM data between engineers from different fields (LOD) representations of each object structure is integrated in a collaborative talking about the same thing, even though composing the scene, and to switch at run- virtual prototyping system developed in they may use different terms; the time between the LODs. This technique the framework of the ESPRIT project possibility of using mock-ups during has multiple drawbacks, both in terms of CAVALCADE (#26285, 1/1998-2/2000). collaborative design sessions among geographically distant partners, a very difficult option when using physical prototypes.

Time-critical Rendering As for all virtual reality applications, virtual prototyping systems have very stringent performance requirements: low visual feedback bandwidth can destroy the illusion of animation, while high latency can induce simulation sickness and loss of feeling of control. Since virtual prototyping tools have to deal with very large dynamic graphic scenes with a complex geometric description, rendering speed is often a major bottleneck. When converted to an adequate accuracy, virtual prototypes often exceed the millions of polygons and hundreds of objects, and this poses important challenges to application developers both in terms of memory and speed. A view of the whole ATLAS Since the complexity of a scene from a Experiment Pit. Comparison given view-point is potentially unbound, between full (top, 1,072,858 triangles) and adaptive resolution meeting the performance requirements of (bottom, 41,487 triangles).

46 ERCIM News No. 44, January 2001 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

All software modules are integrated in an hundreds of independent objects on a choosing the most appropriate methods in-house walkthrough application. We standard graphics PC. In a controlled for different sets of objects based on experimented with the system using environment, it guarantees a uniform, cost/quality considerations and algorithm multiple massive datasets, including the bounded even for widely constraints. All these technologies have digital model of a very large machine changing viewing conditions. The to be integrated in the same rendering (ATLAS) that is under construction at technique does not rely on visibility context. This aspect is often neglected in CERN in Geneva. When completed, this preprocessing and can be readily current research. Integration is made more will be the largest machine for High employed on animated scenes and complex by the fact that dynamic LOD Energy Physics in the world. interactive VR applications. and visibility culling often have different time constraints. The results obtained suggest that our We believe that ultimately a time-critical Links: approach can be usefully employed for rendering system should combine http://www.crs4.it/vvr visual analysis at an early stage in the algorithms such as ours with occlusion design phase of complex engineering or culling, image-based rendering and other Please Contact: architectural models. It can handle scenes acceleration methods. The system should Enrico Gobbetti Ð CRS4 totaling millions of polygons and automatically partition the scene, Tel: +39 070 2796 1 E-mail: [email protected]

Haptic Rendering of Molecular Properties by Aleš Kfienek and Ludûk Matyska

In 1998, a complex research project on human- and extraction of principal features from large data computer interaction was started at the Masaryk sets produced by chemical computations. These university. One of its major streams focused on the methods are based on the interaction via haptic development of new methods for efficient examination interface.

Many interesting chemical properties are Numerous papers were published on the The principal goal of this research is function of force, which is usually method itself and results achieved. design of a haptic interface that allows understood to be a space gradient of the user to modify shape of a molecular potential energy. Visualization techniques Let’s demonstrate the principle on a model (eg, to rotate a particular bond) represent force via 2D or 3D diagrams simple example of the dichloroethan (see while ‘feeling’ the required energy via enhanced by use of colour, ie, user must Figure 1). The molecule is flexible, ie, its force resistance. In the example above the translate visual phenomena to her internal top part can rotate relatively to the bottom user can twist the model feeling resistance force representation. The use of haptic one along the vertical axis. The binding when the atoms get aligned vertically, interface opens a completely new enthalpy (ie, a measurable and maximal force corresponds to perfect dimension, when force (as potential computable quantity corresponding, more alignment of the chlorine atoms. gradient) can be represented and felt or less, to internal energy) depends on the directly, without any unnatural angle of rotation, as shown in Figure 2. A prototype application we presented transformation. Moreover, our physical The difficulty to overcome the rotational recently builds a van der Waals surface experience makes it easy to mentally barrier can be directly related to the force of the modeled molecule and lets the user integrate the force, keeping a notion of necessary to move a point along the curve manipulate a spherical probe of about the overall energy of the system as a in the figure. hydrogen size. The probe can touch the consequence. surface and deform the model by rotating

-21 The idea is currently explored on H computational results of a heuristic H Cl -21.5 systematic search through conformational -22 space of flexible biomolecules. Used -22.5 method of the heuristic search was also -23 developed at Masaryk University, where -23.5 belongs among its well established research areas. The H -24 method was implemented as a production Cl H -24.5 -25 software package, currently consuming -200 -150 -100 -50 0 50 100 150 200 about 30% computational resources of the University’s supercomputing center. Figure 1: Dichloroethan - flexible bond. Figure 2: Dichloroethan - binding enthalpy.

ERCIM News No. 44, January 2001 47 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

the flexible angles if appropriate force is The current implementation therefore • to deploy multiple haptic devices, ie to delivered. This haptic interface is performs the energy minimalisation at achieve more than three degrees of accompanied with 3D visualization of the much slower rate (typically once per freedom of the haptic interface. Also, process. several seconds) and interpolates among we to develop new haptic interface the results. The energy model is also based on circular motion. The implementation uses PHANToM restricted to narrow neighbourhood of force feedback device manufactured by local energy minima in order to further To conclude, we introduced a general Sensable, Inc. The application is reduce otherwise impossibly high concept of delivering information on distributed, runs in UNIX environment, requirements on computational resources. energy of a molecular model (as function using PHANToM Linux driver also of force) via haptic interface. We built a developed at our site. In future we plan to extend our research prototype application that supports the activity in the following ways: concept. However, as the science behind Driving a haptic device requires fast • to build a more general framework for haptic interfaces is very premature, a lot computation of the model - in order to performing the exhaustive energy of effort is still needed to fulfill the create illusion of smooth surfaces etc., the calculation completely off-line and primary expectation, ie, to provide an force has to be recomputed at the rate 1— provide more robust, fast and physically effective interface to huge data sets 3 kHz. However, a convincing dynamic reliable interpolation methods of the emerging from scientific calculation. molecular model requires energy results, - to extend the energy model to Please contact: minimalisation in each step. In general, entire conformational paths. As a Aleš Kfienek Ð CRCIM (Masaryk University) such calculation cannot be done in the consequence let the user flip the Tel: +420 5 4151 2258 available time (300 us - 1 ms). molecule from one conformation to E-mail: [email protected] another

The Virtual Museum of Daily Life in 20th Century Bologna

by Maria Chiara Liguori, Donatella Vasetti, Maria Elena Bonfigli, Antonella Guidazzoli and Alessio Mauri

Museums of Daily Life are dedicated to the a VIrtual MUseum of Daily Life in 20th century Bologna reconstruction of domestic interiors and the and has been conceived as an interactive Virtual displaying of consumer goods and other items related Environment (VE) accessible through the Web and to everyday life in a certain period in time. The MUVI the Virtual Theatre of CINECA, Supercomputing Centre project, partially sponsored by ‘Bologna 2000’, realizes in Bologna.

The MUVI project consists of the have a stronger interest in anecdotes than Daily Life Museums, common in Anglo- development of an innovative interface in historical analysis. Saxon countries, are non-existent in Italy. system (implemented using an XML- The high costs involved in attempting to based markup language) for accessing a The interface enables the user to choose preserve Italy’s enormously rich cultural corpus of information related to which elements are to be visualized heritage mean that there are cost priorities reconstructed virtual environments. 3D during the virtual . In this way, the which make the creation of new museums models of domestic interiors and user will be able to visualize only those very difficult, in particular where this historical information are coded in order virtual components in which he/she is particular type is concerned. to allow the user to personalize the type interested eg architectural elements, of information he/she is given during a objects and their contextualization, However, the use of Virtual Reality and virtual tour in the Daily Life Museum. photos, magazine and TV adverts, TV Interactive Computer Graphics can offer films, recorded speeches, etc. Moreover, a satisfactory alternative solution by: Each user type accesses a tailored version in order to provide the visitor with • overcoming all problems related to of the VE built dynamically and "interesting" background information, space needed for and availability of a interactively on the basis of his/her every 3D object is linked to a historical collection of furniture and/or objects language, profile, particular interests, and description of the resources on which the • allowing a high level of multi-medial level of competence and knowledge. virtual reconstruction is based. In this case involvement through a simplified There is no doubt that different users too, the user can personalize the type of management of music, sounds and (tourists, students, researchers, etc.) have information to be received, obtaining for voices different interests and needs: for example example just the visualization of the • avoiding the staticity, in terms of space a researcher needs a more detailed virtual environment, or simple and time, of a permanent structure. description than a simple visitor, who may descriptions of the 3D objects or more

48 ERCIM News No. 44, January 2001 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

Figure: MUVI, virtual • VRML and Javascript – to develop reconstruction of an Internet acessible 3D models Italian ‘50s kitchen. • Multigen, Performer graphical libraries and Vega development environment – to create an immersive virtual environment accessible at the CINECA Virtual Theatre.

MUVI implements a new way to access a ‘certified’ corpus of 3D models and related historical information - that can be expanded over time – creating a Virtual Museum which can be personalised according to user profiles. The project for the first two rooms, the Italian kitchen and living-room of the Fifties, will end details about the related resources or the characteristics, necessary for a multi- around June 2001. historical hypothesis. platform employment. Links: • The last phase consists of the planning The MUVI project: http://www.cineca.it/muvi/ After studying two sample cases, a kitchen and the development of prototypes for CINECA’s Virtual Theatre: and living-room of the 1950s, the overall the interactive visualization of the http://www.cineca.it/visit/virtualtheatre.html project plan was structured as follows: virtual environments developed. This • The first phase consists of the selection will allow the applicability of the Please Contact: Maria Chiara Liguori, Donatella Vasetti Ð of the sources and the definition of the proposed language to be verified. University of Bologna descriptive language, following a Tel: +39 051 6171920 hierarchical structure that allows for The instruments used in this E- mail: [email protected], [email protected] different customer profiles and access multidisciplinary project, involving platforms. computer scientists, historians and 3D Maria Elena Bonfigli, Antonella Guidazzoli, Alessio Mauri Ð • The second phase consists of the modelers, include: CINECA SuperComputing Centre definition of a methodology for the • XML and Dublin Core - to encode the Tel: +39 051 6171920 development of models – considering information compatibly with standards E-mail: the descriptions of the sources given by already in use in many international {e.bonfigli | a.guidazzoli | a.mauri}@cineca.it the historians – and of the technical museums

Augmented Reality Ð Seeing More by Hellmuth Nordwig

Augmented Reality is the term given by research sponsored by the German Federal Ministry of scientists to a completely new, fascinating technique Education and Research BMBF, working on new which inserts additional information directly into our technology that helps people to see more than what view of the real world. The Fraunhofer Institute for is visible in real life. ARVIKA creates future-oriented Computer Graphics Research IGD in Darmstadt is one solutions for industry, in volving interaction between of the participants in the ARVIKA pilot project human beings and technologies.

In a decade’s time it could be possible to processes. Before the prototype of a new the real view of an object”, explains take a trip in Augmented Reality. “Initial car is built, for example, the engineers put Stefan Müller. This is interesting research work has been under way for the model and its assembly through the wherever 3D digital data is already three years in this field”, reports one of most diverse variants on the computer available as a result of the object being the pioneers: Dr. Stefan Müller, head of using CAD and CAM programs. Rather designed or dimensioned using a the Visualization and Virtual Reality than on-screen, such imaging acquires computer. Some exemplary scenarios department at the Fraunhofer Institute for much more realistic form if presented in indicate potential applications: In the Computer Graphics Research in a ‘CAVE’, the facility installed at the IGD assembly hangar at an aircraft factory: Darmstadt. Virtual, or computer- for three-dimensional projections. The assembly engineer is standing on the simulated reality, is already shaping the wing and routing the cables for the control everyday work of architects, designers “The key aspect of Augmented Reality is surface . The virtual assembly and planners of products and assembly to superimpose virtual worlds on top of instructions show him step by step which

ERCIM News No. 44, January 2001 49 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

‘Augmented Reality’ can Another alternative for the future could be used for training be ‘Virtual Retinal Displays’, in which purposes as well as for the image is projected by laser straight assembly and into the eye. maintenance work. Assembly instructions are displayed on ‘data Stefan Müller describes one of the main goggles’, which allow the challenges for the IGD’s research construction engineer to scientists: “We had to teach the computer look at the real component and relevant in particular how to recognize where the information at the same user is looking”. The best solution turned time. out to be a tiny camera fastened to the data-glasses and sending 25 images per second to the computer. This compares the image information with previously stored orientation points on the object and strand of cable has to be routed where. in practice but the answers to them are, calculates the user’s current position and He has both hands free and does not have unfortunately, normally not very reliable. line of vision. to keep getting down from the wing to Augmented Reality can prevent many a study the layout plan. false move from the outset. Many details still remain to be solved. The range of applications is, however, In the operating theatre: The surgeon has Augmented Reality technology requires simply limitless. More and more products the patient’s medical parameters in view many innovative technological solutions, and facilities are designed on computer at all times and does not need to move his of which some have been developed at nowadays, and it would be beneficial if eyes away to look for data on monitors. the IGD. The key unit is a portable planners could see how their ideas fit into In virtually planned heart or brain computer – not a cumbersome laptop, but the real environment. Augmented Reality operations the surgeon can also reach the an extremely powerful mini-computer, is set to provide the missing link for Man- diseased tissue by precisely the route he which can be attached to the user’s belt Machine-Interaction in the Knowledge- has envisaged. This minimizes the risk – because car mechanics and surgeons Based Society. for the patient. need both hands when they are working. Links Then there are the data-glasses which in http://www.inigraphics.net/publications/ In the office of a town planner or their simplest configuration consist of a brochures/vr_ar.html.en architect: What will be the least obtrusive transparent LCD. In another variant a tiny site for a pedestrian bridge over the screen is attached to the frame of the Please contact: Thames near St. Paul’s Cathedral? Does glasses, from which information is Stefan Müller Ð Fraunhofer-Gesellschaft Tel: +49 61 51/1 55-1 88 this lamp match the planned interior directed to the lens. This lens is actually E-mail: [email protected] design, will it illuminate the required a transparent mirror which guides the area? Such questions frequently crop up information into the sight of the wearer.

Interactive Computer Graphics at Rutherford Appleton Laboratory

by Lakshmi Sastry, Gavin Doherty and Michael Wilson

The two major directions for interactive computer enjoy a game) and to establish a development graphics research at Rutherford Appleton Laboratory methodology for computer graphics Ð as one would are interaction with the graphical representation (how at any other piece of ; can users manipulate graphical representation in real considering efficiency and quality measures. time to gain insight, learn, to reach decisions, or to

Earlier research tried to computerise what cheap and humans are expensive. This that it is in to the user for them to take on was very difficult for humans, and leave has a dramatic effect on interaction their role - which they must clearly the rest to people since they were more between the user and the system. Where understand. plentiful and cheaper. Today, application the interaction used to be there to allow designers take the reverse view: once the user as much control and awareness In applications such as simulations for the data is in the machine, leave the whole of the process as possible, now the training we need to provide an process to the application except when a interaction is only required when there is environment that mimics reality human is absolutely required (for legal something that that the machine can’t do. sufficiently to allow learning in the authorisation etc..), since machines are Then the machine must present the state simulation and promote transfer of that

50 ERCIM News No. 44, January 2001 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

Figure 1: The control room and radar of the EISCAT facility in Tromso, Norway.

Figure 2: The second target station to the ISIS particle accelerator at RAL. learning to the real environment. In an collect and integrate their designs into a 2000, page 21) with Grenoble, Parma, example such as the EISCAT radar single model, but then we need to get the Sheffield and Bath, universities, and control room (see figure 1) we need to designers to see where conflicts arise. For DFKI in Germany, Forth in Greece and simulate the appearance of the radar, the example, where the particle beam passes CNR in Italy. Despite this current activity, building and control machines to provide through a support pillar in the building - a great deal of further work is required to visual cues, as well as the behaviour of clearly impossible in the real world. To incorporate mappings from requirements the devices for interaction with the radar, do this we need to use the graphics to on learning or group decision making and also appropriate feedback. When facilitate group exploration of the overall down to implementable and evaluatable users push switches, or pull leavers they design, and decision making among the geometry, appearance, behaviours, not only need to learn the sequence of designers. We currently do this in a 20 interactions and feedback for sight, sound abstract operations and their seat auditorium providing 3D stereo and touch that use the skills that humans consequences for the radar and control presentation for all, while one user have and machine’s don’t, and even won’t system, but they need to move their hands controls the navigation and interaction given Moore’s law. and body in such a way that the physical with the model. Links: control sequences they learn can be W3C Scaleable Vector Graphics (SVG): transferred to the real control room. They In both these examples we require a clear http://www.w3.org/Graphics/SVG/Overview.htm8 also need feedback to their actions set of requirements of the geometry, ISIS second target station: through sight, sound and eventually touch appearance, behaviour and interaction http://www.isis.rl.ac.uk/targetstation2/ to reinforce this experience. required of the graphics in order to EISCAT radar facility: http://www.eiscat.no/eiscat.html achieve the task objectives, and a INQUISITIVE project: Another application class that we develop methodology to implement and evaluate http://www.cs.york.ac.uk/hci/inquisitive/ to support scientists is that of building that the requirements have been met. In TACIT Project: http://kazan.cnuce.cnr.it/ and experiment prototyping and design one current project with York and Bath TACIT/TACIThome.html review. In this case a new experiment universities that has been previously Please Contact such as a particle accelerator target station reported in ERCIM News (April 2000, Michael Wilson Ð CLRC may be being designed, where the page 49) we have started to develop such Tel: +44 1235 44 6619 building, main accelerator, target, and a methodology. A second project is E-mail: [email protected] each of 20 experiments may each be developing a reference model for designed by different design teams in characterising continuous interaction universities around Europe. We can techniques (see ERCIM News, January

ERCIM News No. 44, January 2001 51 SPECIAL THEME: COMPUTER GRAPHICS AND VISUALIZATION

Conceptual Free-Form Styling on the Responsive Workbench

by Gerold Wesche and Marc Droske

A two-handed styling system for free-form surfaces, tracked, pen-like input device, called stylus. The which runs directly in a table-like Virtual Environment, curves are connected automatically, such that a curve called the Responsive Workbench, has been network develops. Inside the loops, surfaces can be developed at GMD. The user sees the model in 3D computed. Intuitive curve and surface deformation through stereo glasses, which are electromagnetically tools hide the mathematical representation from the tracked. Curves can be drawn directly in 3D using a designer.

Many designers would like to sketch controlling the position and free-form shapes quickly in the orientation of a modelling coordinate phase without using system dynamically or by fixing it. a complex CAD system. Projection- Additionally, the non-dominant hand based Virtual Environments like the can control a virtual transparent Responsive Workbench offer the drawing plane onto which curves are following advantages for sketching projected, or it applies symmetry applications over traditinal desktop planes globally or locally. interfaces: • The three-dimensionality of Future Work perception in combination with 3D The drawback of ‘interacting in the interaction allows an immediate air’, which occurs particularly with understanding of the shape regard to elastic deformations, will be • parts can be localized and selected attacked by integrating a large version directly in space Figure 1: 3D free-form styling on the Responsive of the PHANToM device into the • the large, high-resolution projection Workbench. workbench environment. In contrast plane allows the representation of to this, controlling the smoothing and objects on a scale which corresponds Variational modelling techniques provide sharpening tools described here works to the working region of both hands. an elegant way of manipulating curves very well even without force feedback and surfaces. Attractors of intuitive since the corresponding arm movements Our ambition is to support free-form behaviour can been defined, which are rather indirect. styling capabilities on the Responsive minimize an energy functional and lead Workbench. We provide the designer to fair and pleasant surfaces. Based on Conclusion with intuitive deformation tools for curves this technique, we developed the This article has presented a conceptual and surfaces relying on familiar following intuitive operators, which can design system which supports the intuitive metaphors. Our basic approach is be applied locally or globally to curves transfer of the design intent into virtual constructing curve nets by drawing cubic and surfaces: models. The tools are designed especially B-spline curves directly in space and • smoother: curves can be smoothed out for free-form modelling applications in creating surfaces inside the loops of the locally by pointing to the projection-based Virtual Environments. curve net, see Figure 1. corresponding location directly in The modeler supports the quick input of space. The curve is gradually smoothed surfaces by allowing the user to draw the Features of the Modeler out by moving the pointer along the primal curves of a model directly in 3D. In our system, cubic B-spline curves can curve until the desired shape has been be drawn freely in space. New curves are achieved. In fact, this technique seems User studies have to evaluate the woven into the existing curve network to be similar to applying a flat-iron. usefulness of the proposed techniques to automatically. If a closed loop of curve Currently, the pointing position cannot suggest further improvements. pieces has been formed, a surface patch be changed for surfaces Links: can be created. We are using Catmull- • sharpener: this operation amplifies the Institute for Media Communication: Clark subdivision surfaces. A direct details at the locations pointed to by http://imk.gmd.de approach to modify the curves or surfaces the designer. It works in the same way of the network is to select individual as the smoother Please contact: control points and move them, changing • attractor: segments can be pulled Gerold Wesche Ð GMD Tel: +49 2241 142708 the model locally. It is a very tedious task towards a given direction, while local E-mail [email protected] to bring the curve or surface into the variations of the shape are maintained. desired shape with this technique. Higher level manipulation methods are needed, The non-dominant hand assists the especially in a virtual environment. modelling hand in drawing curves by

52 ERCIM News No. 44, January 2001 RESEARCH AND DEVELOPMENT

Combining Operations Research and Constraint Programming to Solve Real-Life Scheduling Problems by Philippe Baptiste

For a long time, Operations Research and Artificial that we can have the best of both worlds by some Intelligence could be seen as alternative approaches hybrid techniques embeded in a constraint to solve real life optimisation problems such as programming framework. scheduling or resource allocation. It now seems clear

Scheduling, that can be defined as the can be immediately detected. Without are added to the system (eg, instantiating process of allocating scarce resources to propagation, the ‘y ≤ 9’ test could not be a variable to a value, ordering a pair of activities over a period of time, is the key performed before the instantiation of y activities). point of many industrial problems. ‘Good’ and thus no contradiction would be schedules enable to make better decisions detected at this stage of the problem- The separation between constraint about how to commit resources, including solving process. propagation and the other parts of the equipment, capital, people, vehicles, system is a key feature of constraint facilities, etc. The problem of scheduling For complexity reasons, constraint programming. It impacts a lot on the activities is a discrete optimization propagation is usually incomplete. This reusability of the constraint propagation problem and therefore very hard to solve means that some but not all the algorithms in the several applications in practice (most of the scheduling consequences of constraints are deduced. In where similar constraints apply. This problems are NP-Hard). To find good particular, constraint propagation cannot explains the success of commercial and solutions or even solutions that meet all public domain constraint constraints, one has to explore a gigantic programming packages such search space, most often exponential with as ILOG SOLVER, CHIP, the size of the problem. Because a brutal ECLIPSE, CLAIRE and exploration of the search space is not ECLAIR. possible, several techniques have been proposed in the past forty years, including The efficiency of constraint Mixed Integer Linear Programming, based scheduling Branch-and-Bound or more recently applications could be Constraint Programming. drastically improved by using dedicated operations Constraint programming is concerned The behavior of a constraint programming system. research techniques: On the with solving instances of the Constraint one hand operations research Satisfaction Problem (CSP). Informally detect all inconsistencies. Consequently, offers efficient algorithms to solve speaking, an instance of the CSP is tree search algorithms must be problems that however might not be well described by a set of variables, a set of implemented to determine if the CSP suited to be used in practice, and on the possible values (domain) for each instance is consistent or not. The overall other hand constraint programming offers variable, and a set of constraints between behavior of a constraint-based system is algorithms that are more generally the variables. The question is whether depicted on the Figure. The figure applicable, but that might suffer from there exists an assignment of values to underlines the fact that problem somewhat poor performance. It now variables, so that all the constraints are definition, constraint propagation and seems clear that the best of both worlds, satisfied. The interest of this technique decision making are clearly separated: ie, efficient algorithms that can be applied lies in using constraints to reduce the • first, the problem is defined in terms of apply to a wide range of problems, can computational effort needed to solve variables and of constraints be reached by hybrid techniques. combinatorial problems. Constraints are • then, constraint propagation algorithms Following this idea, very good results used not only to test the validity of a are specified. In practice the user of a have been obtained on several industrial solution, as in conventional programming constraint can use applications such as Staff Planning, languages, but also in a constructive mode some pre-defined constraints (eg, Vehicle Production Optimization or Job- to deduce new constraints and rapidly constraints on integers, constraints on Shop Scheduling. detect inconsistencies. sets, scheduling constraints) for which Please contact: the corresponding propagation Philippe Baptiste Ð CNRS, For example, from x < y and x > 8, we algorithms have been pre-implemented Compiègne University of Technology deduce, if x and y denote integers, that • finally, the decision-making process, ie, Tel: +33 3 4423 4952 the value of y is at least 10. If later we the way the search tree is built, is E-mail: [email protected] add the constraint y ≤ 9, a contradiction specified. It states how new constraints http://www.hds.utc.fr/~baptiste

ERCIM News No. 44, January 2001 53 RESEARCH AND DEVELOPMENT

On-line Graph Coloring

by JenΠLehel

Many variants of time tabling, sequencing, scheduling, of the Discrete Structures Group at SZTAKI initiated and allocation problems boil down to the fundamental a natural model for a certain type of these practical problem of partitioning a collection of objects into questions in terms of chromatic . families according to certain requirements. Members

The scheduling of a set of tasks on parallel if there exists an upper bound on its reader can easily check that each graph machines is a basic problem in performance in terms of the on-line is on-line 3-colorable but the hint in the combinatorial optimization. In certain chromatic number of the graphs in the Figure indicates that 4 colors are needed applications, tasks are to be scheduled family. Intuitively, it is a universal on- if the Painter does not know in advance promptly and irrevocably without the line algorithm with reasonable which one is presented by the Drawer, B complete knowledge of the input stream. performance on every member of the or E. It turned out that this elementary The case in which a task must be graph family. example is the best possible one, processed in an uninterrupted fashion (the moreover, there exists an on-line non-preemptive model), the scheduling Against some graph families, for example competitive algorithm using at most four on k machines becomes a question of against the family of trees, the First Fit colors against the family of all on-line 3- coloring intervals or interval graphs on- coloring algorithm is an on-line colorable ghraphs. line with k colors. An equivalent scenario competitive algorithm. However, this of dynamic storage allocation due to property does not hold in general: against Many results in the theory of on-line graph Marek Chrobak and Maciej Slusarek, the family of on-line 3-colorable bipartite colorings can be formulated in terms of motivated us to introduce the notion of graphs the First Fit algorithm is not on- on-line competitive algorithms. Some of on-line coloring and on-line chromatic line competitive. these results demonstrate that against number of graphs or, more generally, certain families of graphs these algorithms families of graphs different from interval An example shows that an on-line are competitive in a stronger sense graphs. The research of the author with competitive algorithm against the family (compared with off-line ones). The main András Gyárfás (SZTAKI) and Zoltán of 3-colorable bipartite graphs must use open problem concerning on-line Király (Loránd Eötvös University, at least four colors: just consider the pair competitive algorithms is to decide Budapest) in this area over the last decade of graphs B and E in the Figure. The whether there exists an algorithm like this contributed to the development of against the family of on-line k-colorable classical graph theory with several graphs, for any fixed k. The answer is not noteworthy results and challenging open even known for special graph families like problems. Furthermore, it seems to have the families of bipartite graphs, chordal a certain impact on the theory of on-line graphs or even chordal bipartite graphs. algorithms, a recently outgrowing branch of combinatorial optimization. In addition to contributing to the development of recursive , The on-line chromatic number of a graph our research inspired some results of G is usually defined through a two-person experts eg Hal Kierstead and William game. The Drawer presents the vertices Trotter. Their earlier deep results on of G in some order together with partially ordered sets have got a new adjacencies to vertices already given. The interpretation in terms of on-line Painter assigns a legal color (positive partitioning or coloring algorithms of integer) to the current vertex given. The graphs. The concept of on-line algorithms aim of the Drawer is to force the Painter combined with graph theory turned to be to use as many colors as possible and the a steady source of challenging new aim of the Painter is to use as few colors problems and led to a number of as possible. The common optimal value interesting graph theory results obtained is the on-line chromatic number of G. The by several competitive research groups. most familiar on-line coloring algorithm Our group is open for any further is the First Fit (alias Greedy) algorithm discussions. Drawer’s strategy is indicated when playing which assigns the smallest proper color on the pair of graphs B and E. Possible Please contact: to the current vertex. outcomes of Painter’s on-line coloring are JenŒ Lehel Ð SZTAKI shown on the left after the first 4 or 5 moves, and University of Louisville followed by Drawer’s corresponding moves An on-line coloring algorithm is called Tel:+ 36 1 4665 644 on the right to conclude the game by forcing E-mail: [email protected] on-line competitive against a graph family 4 colors.

54 ERCIM News No. 44, January 2001 RESEARCH AND DEVELOPMENT

Boosting Algorithms for Automated Text Categorization by Fabrizio Sebastiani, Alessandro Sperduti and Nicola Valdambrini

As part of its activities, and in on the construction of tools for the automatic or semi- collaboration with the Department of Computer automatic labeling of texts with thematic categories Science of the University of Pisa, IEI-CNR is working or subject codes.

Text categorization (TC) uses machine effectiveness and a strong theoretical scores of the best terms are markedly learning (ML) techniques to build grounding in computational learning different from each other and from the automatic text classifiers, ie programs theory. In the boosting method, the S worst ones, in the last iterations the capable of labelling natural language texts classifiers that make up the committee are differences among scores are very small. with thematic categories from a not trained in a conceptually parallel and In AdaBoost.MH(KR) we choose, at each predefined set. TC has many applications, independent way, as in other committee- iteration, K top-ranked terms that would including automated document indexing based methods, but sequentially. In this be good candidates for selection in the for Boolean systems, way, when generating the i-th classifier next K AdaBoost.MH(R) iterations. We adaptive document filtering and spam e- the learner can take into account how the can thus build a committee composed of mail blocking, Web page classification previously generated classifiers perform S sub-committees at a computational cost under hierarchical directories, and even on the training documents, and can force comparable to that required by author identification for documents of the i-th classifier to concentrate on those AdaBoost.MH(R) to build a committee controversial paternity. In TC a general training documents where the previously of S classifiers. In fact, most of the inductive process (called the ‘learner’) generated classifiers performed worst. computation required by the boosting automatically builds a classifier by process is devoted to calculating the term learning the characteristics of the Each classifier is a one-level , scores, and by using only the top-scoring categories from a ‘training set’ of ie tests for the presence or absence of a term AdaBoost.MH(R) exploits these preclassified documents. Once a classifier term in a document, depending on hard-won scores only to a very small has been built, its effectiveness (ie its whether a positive or negative extent. capability to take the right categorization classification decision is taken. The term decisions) may be tested by applying it on which this decision hinges is generally In all our experiments (conducted on the to a ‘test set’ of preclassified documents different for each of the S classifiers and standard Reuters-21578 benchmark), and checking the degree of is chosen each time, according to a AdaBoost.MH(KR) achieved better correspondence between the decisions of complex error-minimization policy. The effectiveness and much better efficiency the automatic classifier and those encoded key step of this policy consists in scoring than AdaBoost.MH(R). For instance, in in the test set. each term by a scoring function, and one experiment the effectiveness achieved picking the term with the smallest score. by AdaBoost.MH(KR) with 300 iterations A wealth of different ML methods have The scoring function adopted by the was not obtained by AdaBoost.MH(R) in been applied to TC, including boosting method is extremely complex, 16,000 iterations. AdaBoost.MH(KR) is probabilistic classifiers, decision trees, and computationally onerous. As this more effective than AdaBoost.MH(R) regression methods, batch and computation has to be performed for each because of the latter’s ‘greedy’ approach incremental linear methods, neural of the terms and this has to be iterated S when selecting the best classifier for the networks, example-based methods, and times, boosting is a computationally final committee. This approach does not support vector machines. Recently, the expensive method (in typical TC guarantee optimality of selection, and ‘classifier committees’ method has gained applications both the number of terms and AdaBoost.MH(R) is unable to search the popularity. The underlying idea is that, the number S of iterations are in the tens classifier space. On the contrary, given a task that requires expert of thousands). AdaBoost.MH(KR) allows the designer knowledge, several independent experts to explore, at least partly, the space of are better than one if their individual We have devised a new method, called classifiers by setting its internal judgments are appropriately combined. AdaBoost.MH(KR), based on the parameters (eg number of classifiers in In TC, this means applying several construction of a sub-committee of each committee) to different values or different classifiers to the same task of classifiers at each iteration of the boosting according to different policies. deciding under which category (or process. Instead of using just the best Links: categories) a document should be term, as the standard AdaBoost.MH(R) http://faure.iei.pi.cnr.it/~fabrizio/Publications/ classified, and then combining their algorithm does, we select the best K CIKM00/CIKM00.pdf outcome appropriately. terms, and generate K classifiers, grouping them in a sub-committee. This Please contact: Boosting is a method for generating idea comes from the observation that, Fabrizio Sebastiani Ð IEI-CNR Tel: +39 050 3152 892 classifier committees which is while in the first iterations (ie in the E-mail: [email protected] characterized by state-of-the-art generation of the first classifiers) the

ERCIM News No. 44, January 2001 55 RESEARCH AND DEVELOPMENT

Pattern Formation in Electric Discharges

by Ute Ebert

If a material that normally does not conduct electricity, particular in so-called streamer and barrier discharges. is exposed to strong electric fields, it can develop The results can be used for combustion gas cleaning, complex patterns of conducting areas, through which ultrafast power semiconductor switches or for flat it ‘discharges’ the field. A group at CWI studies both computer screens. fundamental and applied aspects of such patterns, in

Electric discharges (sparks) appear on a variety of length scales: in nature between a cloud and the ground (Figure 1 shows man-made lightning across about one meter), in the sparkplug of a car engine, or in computer chips. Sparks or their precursors, the so-called streamers, are also used for, eg, combustion gas cleaning (see Figure 2), or water purification. Streamer-like discharges in doped layered semiconductors can be used for ultrafast switching of high power electric circuits.

Stationary discharges are also possible, Figure 3: Generic instabilities in ac barrier the best known example being the neon glow discharges with wide coplanar tube. Since a discharge is a nonlinear electrodes seen from above (I. Brauer and process that easily can form patterns, the H.-G. Purwins, Univ. Münster). Here areas homogeneous and stationary glow of the Figure 1: Spark demonstration at the with homogeneous glow and areas with ‘Deutsches Museum München’. spontaneous glow filamentation coexist. neon tube is by no means trivial, and stays a continued challenge in the development of new energy saving and special purpose electric field is high enough. However, ultrafast semiconductor switches, the light sources, eg, at Philips. Figure 3 conducting matter can redistribute its onset of pattern formation limits the range shows generic instabilities that can occur space charges and hence modify the of applicability. in such glow discharges. Particular effort externally applied electric field. is presently devoted to the development Depending on the spatial distribution of There is a surprising universality in the of so-called plasma display panels: flat the ionized areas, the ionization reaction set of partial differential equations and bright computer or TV screens then is enhanced at some places and describing discharge phenomena in operating essentially like an array of suppressed at others. This nonlinear different materials. They are of rapid, microscopic neon tubes. dynamical coupling of ionization and comparable simplicity and generality as, electric field accounts for the spatio- eg, the Navier-Stokes-equation in The enormous increase of electric temporal patterns observed in discharge hydrodynamics. Nevertheless, despite the conductivity in a discharge is due to the experiments. In some applications like enormous economical importance of multiplication of free charge carriers by combustion gas processing or spark plugs, discharge technologies, very little is an approximately local ionization pattern formation is the operating mode, known about basic solutions of the reaction. It persists as long as the local while in others like light sources or fundamental equations. However, recent developments in the theory of nonlinear partial differential equations and nonlinear dynamics promise that progress can be made in unravelling these features.

At CWI the basic laws of propagation of the ionized streamer channels — that can be seen in Figure 2, and that have preceeded and determined the spark in Figure 1 — are studied analytically and Figure 2: Temporal evolution of a streamer discharge within nanoseconds in a combustion gas numerically. A basic insight is a strong processor at the TU Eindhoven [P.P.M. Blom and E.J.M. van Heesch]. The picture is taken on mathematical analogy to patterns formed the axis of the electrode wire, and the maximal radius of the discharge is about 10 cm. in viscous fingering or in the growth of

56 ERCIM News No. 44, January 2001 RESEARCH AND DEVELOPMENT

crystals from undercooled melts. Another who received the prize for the invention structure. Here mathematical analogies, study concerns ultrafast high-power of the semiconductor GaAs laser last fall. eg, with temperature driven convection semiconductor switches that operate in a A third project concerns the basic in fluids or with chemical systems are streamer-like manner. Applications are instabilities in barrier discharges as in exploited. found in radar and laser technology, for Figure 3. Barrier discharges consist of a Please contact: example a possible new generation of sandwich structure of electrode, resistive Ute Ebert Ð CWI GaAs pulsed lasers. This is a collaboration layer, glow discharge and electrode, Tel: +31 20 592 4206 with the Ioffe-Institute in St. Petersburg driven by dc or ac voltage. The E-mail: [email protected] that is internationally leading in the mechanism of pattern formation is only http://www.cwi.nl/~ebert production of these switches, and is partially understood, and we work on headed by Nobel laureate Z.A. Alferov quantitative predictions of its onset and

Design of Efficient Input/Output Intensive Data Mining Applications by Ranieri Baraglia, Domenico Laforenza, Salvatore Orlando, Paolo Palmerini and Raffaele Perego

The goal of Data Mining is to discover knowledge minimizing their effects. At CNUCE-CNR we are hidden in data repositories. This activity has recently studying how the architectural features of modern attracted a lot of attention. Because of the huge high performance computers and operating systems datasets to be accessed, most data mining algorithms can be exploited when designing efficient data mining must consider the I/O actions carefully, hiding or algorithms that can cope with huge datasets.

High energy physics experiments produce Computing Systems, which can extend categories of algorithms, with respect to hundreds of TBytes of data, credit card the range of applicability of DM the patterns of their I/O activities: Read companies hold large databases of algorithms, without changing the and Compute (R&C) algorithms, which customer’s transactions, web search conceptual limitation. use the same dataset at each iteration, and engines collect web documents Read, Compute and Write (RC&W) ones, worldwide. Regardless of the application Many DM algorithms require a which, at each iteration, rewrite the field, Data Mining (DM) allows to ‘dig’ computation to be iteratively applied to dataset to be used at the next step. A into huge datasets to reveal patterns and all records of a dataset. In order to typical reason for writing a new dataset correlations useful for high level guarantee scalability, even on a serial or is to ‘prune’ the original one. interpretation. Finding clusters, a small scale parallel platform (eg a association rules, classes and time series workstation cluster), the increase in the The first class includes Clustering are the most common DM tasks. All I/O activity must be carefully taken into algorithms, such as ‘k-means’, or require the use of algorithms whose account. We have recognized two main Frequent Set Counting (FSC) algorithms, complexity, both in time and in space, grows at least linearly with the database size.

In recent years much R&D has been focused on the design of hardware and software systems that can cope with the growth in the dimensions of the data. Nevertheless, the amount of physical memory of modern computers is still in orders of magnitude lower than the size of many databases. Various strategies can be applied to address this problem. One approach investigates the development of new algorithms that reduce the need for data, either by exploiting sampling techniques or by limiting direct access to the database. These methods usually introduce some inaccuracy or useless CPU overhead. Another possibility is the The result of an experiment implementing an R&C problem. Total execution time and I/O+idle exploitation of High Performance time are plotted, varying the computational grain.

ERCIM News No. 44, January 2001 57 RESEARCH AND DEVELOPMENT

such as ‘Apriori’, while the second class the in-core counterparts. This is mainly by partitioning data among the covers some variants of ‘Apriori’, like the because data locality can be exploited machines of the cluster. This makes it Direct Hash and Prune (DHP) algorithm, and I/O actions can be overlapped with possible to scale-up the system since or the C4.5 algorithm for classification. computation. The figure refers to a storage and computation power are synthetic kernel code inspired by this increased, and to exploit a larger I/O Modern operating systems (OS) provide class of DM algorithms. In this kernel, bandwidth since multiple disks effective mechanisms to access secondary we scan a file of 256 MB sequentially, interfaced with multiple buses are storage units. While highly local problems reading blocks of 4 KB. We artificially employed. We also developed dynamic can exploit OS caching and prefetching varied the granularity of the techniques, such as load balancing and policies, the OS may fail to efficiently computation performed on each block adaptive modification of the degree of handle data movements through the levels of data. When the total CPU time is parallelism, to take into account system of the memory hierarchy. The greater than 20 s, the I/O and idle heterogeneity and improve system employment of ‘Out-of-Core’ (OOC) waiting times become negligible due to utilization. techniques which explicitly take care of the prefetching OS activities. We data movements is thus often a must for discovered that most of the interesting Our next goal will be to extend the DM performance reasons. Another important DM problems belonging to the R&C set of algorithms tested against the general OS feature is time-sharing among class can be solved by algorithms results found so far. We are now studying processes. Multi-threading can be used to characterized by high granularities; this a classification algorithm for building overlap I/O actions with useful means that our OOC techniques are decision trees and algorithms for data computations. able to completely hide I/O overheads. mining on the Web. We intend to build a • Multi-threading used as a general set of methodologies for DM I/O We have demonstrated that it is possible technique to hide I/O waiting time may intensive algorithms, characterized by to take advantage of such features in order conflict with prefetching, another application independence and run time to design efficient OOC DM algorithms technique independently adopted by the adaptability. The methodologies should that hold all the data on disk, and that OS to reduce the impact of I/O when be implemented as a general tool in an these algorithms can be effectively scaled files are accessed sequentially. We Application Programming Interface. by exploiting High Performance developed techniques to take advantage Links: Computing Systems, such as a cluster of of both multi-threading and prefetching http://miles.cnuce.cnr.it/~palmeri/datam Symmetric Multiprocessors (SMPs). The when implementing OOC DM main results are summarized in the algorithms. Please contact: following points: • On clusters of SMPs, data parallel Paolo Palmerini — CNUCE-CNR • In R&C DM algorithms we find that approaches have been used to Tel: +39 050 315 2967 E-mail: [email protected] OOC techniques perform better than parallelize these OOC DM algorithms

Signware Ð A Learning Platform in Biology for Pupils with Hearing Difficulties

by Kathrin Bolits

The project Signware is developed in conjunction with regular curriculum, at secondary high school level for a diploma thesis at GMD and the University of Cologne pupils with hearing difficulties. The content subject to implement concepts and realisation of a learning of the learning tool is biology with the main theme platform. Signware is aimed to form a part of the focusing on ‘blood’ in general.

Signware in a nut shell is an interactive limited abilities through audio they have to keep their attention on three software tool which enables pupils with perceptions, the visualization as well as aspects during a discussion: hearing difficulties to learn about ‘blood’ cognitive, motoric and social-emotional at large. The main element of Signware aspects are the constituting the essential • language perception which is disturbed is the visualization of the German Sign parts of Signware. according to the degree of the hearing Language (DGS), coupled with computer damage animations to motivate the user. It is well To understand the communication • language processing (meaning the known that interactivity greatly enhances problems of individual with hearing cognitive aspect about what to say) the level of reception. To cater for the difficulties, it is essential to know that • language production (meaning the special requirements of the users and their motoric aspect of talking).

58 ERCIM News No. 44, January 2001 RESEARCH AND DEVELOPMENT

Listeners with adequate hearing receptions only require to concentrate on language processes, whereas people with hearing difficulties need much more effort to follow and to participate in a communication. Hence for the latter group a higher degree of concentration is required when comprehending some discussions. Figure1: of blood - starting page. Figure 2: blood drop which guides the user When teaching pupils with hearing through the programme. difficulties, it is generally recommended to provide some additional information user through different sections of the tool. situations. For instance in the case of such as visual representations to enhance In this way, the concentration and hence ‘blood’, an animated sequence defines comprehension. the comprehension of the learner in how blood could drop from a finger tip, question increases dramatically. when it is cut. Structure The programme is designed in such a way Practical Example Further Developments that the learning modules are based on To test the efficiency of Signware, some The present version of Signware some logical didactically structured units. practical examples are also available in comprises of three completed learning The latter are divided through a number the programme. One such example is modules, namely: ‘Introduction to the of single visual sites which are created simply say to describe the word ‘blood’ Programme’, ‘Blood Channels’ and some through the integration of data formats to the user. By clicking on a respective extracts on ‘Blood Functionalities’. such as graphics, videos and animations. button identified by the word, one may Further development in the part of Moreover the visual sequences are not only receive information about ‘blood’ multimedia representation of the tool is coupled with spelled words of the German in general but also observe the German underway. This include visitation of more Sign Language which may be activated Sign Language spelling of the word. scientific signs like ‘erythrocyte’ and through interaction. ‘carbon dioxide’ to facilitate using the When clicking on a drawn pencil it is German Sign Language. The content is prepared using didactical possible to receive further information in Please contact: principles of reduction and reproduction. the forms of graphic and text on the Kathrin Bolits Ð GMD Guiding through the programme is subject extending the pupil’s knowledge Tel: +49 2241 14 2132 catered for using a cartoon character in and capabilities. In addition it is possible E-mail: [email protected] the form of a blood drop, which leads the to define words’ meanings in appropriate [

Subscription Form Name:

If you wish to subscribe to ERCIM News Organisation/Company: or if you know of a colleague who would like to receive regular copies of ERCIM Address: News, please fill in this form and we will add you/them to the mailing list.

Send or fax this form to: Post Code: ERCIM NEWS BP 105 City: F-78153 Le Chesnay Cedex Fax: +33 1 3963 5052 E-mail: [email protected] Country

Data from this form will be held on a computer database. By giving your email address, you allow ERCIM to send you email. E-mail:

You can also subscribe to ERCIM News and order back copies by filling out the form at the ERCIM web site at http://www.ercim.org/publication/Ercim_News/

ERCIM News No. 44, January 2001 59 EVENTS

SPONSORED BY ERCIM MFCS 2000 Ð Mathematical Foundations of Computer Science

by Branislav Rovan

The international symposium Mathematical It was organized under the auspices of the Minister Foundations of Computer Science (MFCS) is one of of Education of the Slovak Republic Milan Ftacnik by the top yearly events in theoretical computer science the Slovak Society for Computer Science and the in Europe. The silver jubilee 25th MFCS took place on Faculty of Mathematics, Physics, and Informatics of August 28 - September 1, 2000 in Bratislava, Slovakia. the Comenius University in Bratislava.

This series of symposia was initiated in Committee together with many other Over 120 participants from 20 countries 1972 in Poland and they are held leading research institutions in Europe and (USA, Canada, Israel, South Africa, and alternatly in Poland, Slovakia, and the Northern America. The Proceedings of most European countries) took part at Czech Republic. The main goal of the the symposium were published by the MFCS 2000. The symposium was series of symposia is to stimulate research Springer-Verlag (http://www.springer.de) supported by the European Association for in all branches of theoretical computer as Vol. 1893 of the Lecture Notes in Theoretical Computer Science (EATCS science. Its broad scope facilitates Computer Science series. Three http://www.eatcs.org), ERCIM, and the interaction among researchers who do not specialized workshops were organized on Slovak Research Consortium for usually meet at specialized conferences. the days preceeding and following the Informatics and Mathematics (SRCIM MFCS: Algorithmic Foundations of http://www.srcim.sk). Telenor Slovakia The scientific program of MFCS 2000 Communication Networks coordinated by supported the symposium by providing the consisted of invited talks by A. Gordon Marios Mavronicolas (Nicosia), New Internet connection to the conference site (Cambridge), J. Davenport (Bath), R. Trends in Formal Languages vs. and by hosting the symposium web page. Grosu (Philadelphia), L. Hemaspaandra Complexity coordinated by Klaus W. Links: (Rochester), G. Italiano (Rome), J. van Wagner (Wuerzburg), and Prague Symposium web page http://www.mfcs.sk Leeuwen (Utrecht), M. Vardi (Houston), Stringology Club Workshop coordinated S. Zaks (Haifa) and of 57 contributed talks by Borivoj Melichar (Prague). Please visit Please contact: selected from the 147 submitted the symposium web page at Branislav Rovan Ð SRCIM contributions by the 25 members of the http://www.mfcs.sk for more details on Tel: +421 7 6542 6635 E-mail: [email protected] Program Committee. Five ERCIM the program and attached workshops. institutes were represented on the Program

SPONSORED BY ERCIM EUROGRAPHICS 2000 Conference Report

by David Duce

The Eurographics 2000 conference, sponsored by Geneva in 1980. The 1990 conference was held in ERCIM, was held in Interlaken, Switzerland, from 21 Montreux, so it was very appropriate to return to to 25 August 2000. The event was the 21st annual Switzerland after a further interval of 10 years. Eurographics conference, the first being held in

The theme of the conference was ‘Global SIGGRAPH and the latter are published and a further 24 in short presentation Challenges of the Information Society: A in book form by Springer-Wien. sessions. Six state of the art reports were Web of Opportunity’ with sub-themes on included in the programme and provide modelling, visualization and collaboration. The conference itself spread over three very substantial reviews of areas such as The conference proper was preceded by days. There were three invited papers. recent advances in visualization of two days of tutorials on a wide variety of The first by Robert Cailliau was a volumetric data, shadow computation: a topics in the visualization and modelling thought-provoking account of his unified perspective, etc. areas, plus two workshops: Graphics as a user of 2000 organised in association graphics. Roger Hubbold gave the second The conference has a number of paper with ACM SIGGRAPH and Computer on collaboration in virtual environments award categories. The best student paper Animation and Simulation (CAS 2000). and Daniel Thalmann gave the third on award went to: ‘Real Time, Accurate Proceedings of the former will be available virtual humans – ten problems not Multi-Featured Rendering of Bump through Eurographics and ACM completely solved. A total of 52 papers Mapped Surfaces’ by M. Tarini, P. were presented in the full paper sessions, Cignoni, C. Rocchini, and R.

60 ERCIM News No. 44, January 2001 EVENTS

Scopigno.3rd place overall went to work that would be worthy of the John The full conference papers are published ‘Multiresolution Shape Deformations for Lansdown award. John was known for in the journal Computer Graphics Forum, Meshes with Dynamic Vertex the way he saw things from a different 19(3). Copies of all conference Connectivity’ by L.P. Kobbelt, T. angle to most of us, often bringing new publications are included on the Bareuther, H.-P. Seidel, in 2nd place was insights by an off-beat approach. It was conference CD-ROM, for details please the paper by Tarini et al. and the Gunter this element of the interaction that the contact the Eurographics Association Enderle Award for best paper went judges felt was generally not present in (contact details are at the web site). toExternal Memory View-Dependent the submissions, but some works Simplification’, by J. El-Sana, Y.-J. submitted by Fariba Farshad of the Participants from 29 countries attended Chiang. The full text of the award London Institute (UK) got very close to the event. There was very good winning papers is publicly available at the judges’ requirements, so a second participation from students, which gave the Eurographics Association’s web site: prize was awarded to ‘Celebrating a conference an all-important youthful http://www.eg.org/. Creativity’. The CD-ROM is a showpiece atmosphere! ERCIM sponsorship of the for awards for schoolchildren and event was very important to us, and is A new prize was launced this year, the their teachers, with prize-winning works very gratefully acknowledged. John Lansdown Prize, given in memory in various categories being displayed. of Professor John Lansdown a key figure There is a playful, game-like interface It is not too early to start preparing a paper in computer graphics and computer art that is engaging to users, encouraging for the Eurographics 2001 conference, who died in January 1999. The judges them to seek out the different works which will be held in Manchester, UK. were looking for an interactive web site displayed. Details of the competition Further information: or CD-ROM that works technically and being organised with Eurographics 2001 http://www.eg.org/ artistically and that is innovative in its use can be found at the web site together with of interaction. All judges being active in a fuller account of the criteria by which this field, they had in mind the quality of competition entries will be judged.

SPONSORED BY ERCIM FORTE/PSTV 2000 by Tommaso Bolognesi

The 20th anniversary of the IFIP WG6.1 Joint was celebrated in Pisa, Italy, in October with FORTE International Conference on for XIII / PSTV XX. There were over 90 participants in this Distributed Systems and Communication Protocols very successful ERCIM sponsored event.

The programme committee selected 22 Several presentations addressed the biology, social sciences, arts and papers for presentation at the Conference analysis of communication protocols, humanities, music. After two decades, out of the more than 60 submitted. The with some of them paying particular FM’s are perhaps ready to spread out of central themes were the theory and attention to multicast protocols. Other their native territory and invade new practice of Distributed System papers dealt with the specification, exciting areas of research, for a wider Verification (with Model Checking implementation and testing of hardware exploitation of the huge intellectual playing a predominant role), and Testing. systems. The issue of formal semantics investment behind their definition. That the early pioneering phases in the was also covered, but to a lesser extent development of Formal Methods (FMs), than in past editions of the Conference; a A few copies of the Conference with their misconceptions and myths and further indication of the shift of emphasis Proceedings (published by Kluwer their over-optimistic applications to toy from the definition to the application of Academic Publishers) and of the Tutorial examples and over-skeptical views about FM’s. Three invited speakers - Rocco De Notes are still available; please e-mail scalability to industrial cases, are Nicola, Fausto Giunchiglia, and John Grazia Carrai at Consorzio Pisa Ricerche essentially over was shown by a number Rushby - gave talks on Verification and if interested ([email protected]). of papers that reported on successful Security protocols and there were also Please contact: experiences in specifying and verifying two tutorial/advanced seminar tracks. Tommaso Bolognesi Ð IEI-CNR real distributed systems and protocols. Tel: +39 050 3152797 The Conference also indicated that testing On the first day, a multidisciplinary E-mail: [email protected] theory and applications remain Satellite Workshop — ‘Formal Methods remarkably healthy, showing that the idea Elsewhere’ - was devoted to applications Diego Latella - CNUCE-CNR Tel: +39 050 3152982 that the adoption of FMs would of FM’s to areas other than E-mail: [email protected] eventually eliminate the need for testing communication protocols and distributed is still far from becoming a reality systems, such as physics, chemistry,

ERCIM News No. 44, January 2001 61 EVENTS

SPONSORED BY ERCIM AD2000 Ð The Third International Conference on Automatic Differentiation: From Simulation to Optimization

by Uwe Naumann

111 people from all over the world, including Russia, have created a momentum. Apart from having Brasil, China, and Japan, attended AD2000, the third motivated further research and development they in a series of workshops devoted to Automatic gathered researchers from different fields, from Differentiation. The two previous conferences mathematics to scientific programming and to (Breckenridge 1991 and Santa Fe 1996, both USA) software engineering.

The domains covered by AD2000 parameters settings. With the help of AD Systems) and academia (University of reflected these different horizons. techniques, this trial and error approach Western Ontario, Cananda, Argonne Mathematical aspects included the could be replaced by a more efficient National Laboratory, USA) with strong differentiation of iterative processes, and optimization with respect to modes, design, participation of the audience underlined the analysis of non-smooth code. The use and control parameters. AD tools have also the importance of the development of a of adjoints in optimization and the grown in variety, power and complexity. commercially supported AD tool. propagation of rounding errors were Their presentations and general discussions looked at from the scientific programming turned out to be highly profitable. The proceedings of AD2000 will be point of view. Implementation problems, published by Springer next year. Topics such as the complexity of the different The objective of this third conference was covered will include: Sensitivities for differentiation modes, memory to continue the work started in the previous ODEs and Optimal Control; Optimal management problems, and static compile- meetings, and present recent developments Design via Adjoints in PDEs; AD Tools time analysis to improve differentiated to the audience. Special emphasis was put and Environments; Higher Order programs were also discussed. on the relationships and synergies between Derivatives; Diverse Applications. AD tools and other software tools such as Links: Since the last meeting in 1996, research and parallelizers. Also, the Conference website: has progressed considerably. AD is being maturation of AD tools motivated a special http://www-sop.inria.fr/tropics/ad2000/ applied to larger and larger programs, for session on the use of AD in an industrial example for optimum shape design. In context. A panel discussion between Please contact: many such applications, modelling was several invited speakers from industry Uwe Naumann Ð University of Hertfordshire E-mail: [email protected] restricted to simulations at various (NAG, Sun Microsystems, Frontline

Joint ERCIM Workshop on Databases and Environmental Modelling

by Brian Read and Thomas Lux

The ERCIM Environmental Modelling and the Database Cybernetics Institute of the island’s new University. Research Groups combined to hold a joint Workshop With twenty-three researchers from twelve institutions on 27 to 29 September 2000. It was on Gran Canaria participating, the seventeen speakers addressed the at the generous invitation of Professor Roberto topic of the application of models and databases for Moreno-Díaz, Director of the Technology and environmental research.

The location of the Workshop was the and soil pollution, use of remote sensing expert systems), image and visual historic Casa de Colón in the old part of data for environmental management, and processing, data uncertainty, validation, the city of Las Palmas. The presentations the improvement/parallelisation of integration and visualization. and discussions focused on perspectives numerical algorithms in environmental to advance the dialogue among researchers models. Information systems technologies Besides the regular presentations of the working on environmental modelling and covered were data mining, data working group members, two additional database research, aiming to bring new warehousing, multidimensional data and sessions have to be emphasised. The first, ideas into these fields by intelligent knowledge discovery in environmental very interesting one presented the current systems approaches and GRID computing. databases, web and semi-structured data research topics of the host Instituto management, geographic information Universitario de Ciencias y Tecnologías Crucial environmental research issues systems, decision support systems, Cibernéticas of the Universidad de Las addressed included: modelling of air, water knowledge-based systems (neural nets, Palmas de Gran Canaria as a new member

62 ERCIM News No. 44, January 2001 EVENTS

of the Working Group Environmental current tasks of the DECAIR project, an Links: Joint Workshop: http://www.first.gmd.de/ Modelling. A second, valuable session EC R&D collaboration also initiated by applications/ercim7.html presented by Guy Weeks (EC, DG IST), the Environmental Modelling Group. Working Group Environmental Modelling: focused on the IST Programme within the http://www.first.gmd.de/applications/em.html Fifth Framework Programme of the Overall it was an enjoyable meeting, Database research group: European Commission. Encouraged by benefiting from the interaction of the two http://www.ercim.org/activity/wg/EDRG/ DECAIR Project homepage: the latter, new project ideas were ERCIM Groups within a traditional social http://www-air.inria.fr/decair/ discussed. These lead to a new project programme. During the Workshop, a proposal of the Working Group formal meeting with the Island’s President Please contact: Environmental Modelling that was and the Minister for Science and Achim Sydow Ð GMD submitted to the fourth Call for Proposals Education provided the opportunity to Tel: +49 30 6392 1800 E-mail: [email protected] of the IST Programme by the end of thank them for their hospitality and to October deadline. In addition, the promote the value of ERCIM in Keith Jeffery Ð CLRC opportunity was taken during the integrating Las Palmas within Europe. Tel: +44 1235 44 6103 workshop for a working meeting on E-mail: [email protected]

First Results of the Cross-Language Evaluation Forum by Carol Peters

More than twenty research groups participated presented at a Workshop in Lisbon, 21-22 September, actively in the first campaign of the Cross-Language immediately following the fourth European Conference Evaluation Forum (CLEF 2000). The results were on Digital Libraries - ECDL 2000.

The Cross-Language Evaluation Forum The results of the activity were presented different groups in their experiments were (CLEF) is an activity of the DELOS during a two-day workshop on Cross- published in the Working Notes printed Network of Excellence for Digital Language Information Retrieval and by ERCIM as part of the DELOS Libraries. The goal of CLEF is to provide Evaluation, 21-22 September, Lisbon, Workshop series and distributed at the an evaluation infrastructure and Portugal. The first day, attended by nearly Workshop. benchmarking facilities for the testing and sixty participants, was open to all those tuning of monolingual and cross-language interested in the area of Cross-Language Details of the two-day Workshop, information retrieval systems operating Information Retrieval (CLIR). The including slides of the presentations and on European languages. morning was dedicated to the presentation copies of the Working Notes, can be found of invited papers and discussion sessions on the CLEF Web site. The Proceedings The first evaluation campaign offered four on CLIR research related issues. The will be published by Springer in their evaluation tasks designed to test afternoon addressed the topic of CLIR Lecture Notes for Computer Science multilingual, bilingual, domain-specific, system evaluation: the current situation series. The volume will include a record and monolingual (non-English) IR and future developments. The objective of the talks given on Day 1 and of the systems. The main task consisted of was to promote a discussion on ‘What is experiments and results of the CLEF querying a multilingual corpus of needed to improve CLIR system evaluation campaign presented on Day 2. newspaper documents in four languages performance’ and ‘How can evaluation All papers will be revised and extended (English, French, German and Italian) and campaigns assist in this’. The goal was to with respect to the preliminary draft submitting the results in a ranked, merged identify the actual contribution of version list. Any one of eight European languages evaluation to system development and to could be used to query this collection. determine what could or should be done The programme and schedule for CLEF in the future to stimulate progress. 2001 is to be found on the CLEF Web site. Eight North American and twelve The agenda will be similar to that for 2000 European research groups, mainly from The second day was restricted to but with the addition of more languages academia – a few from industry, managed participants in the CLEF 2000 evaluation to the multilingual text collection to submit results within the 1 July campaign. The results of the evaluation (Spanish, Dutch and perhaps Greek). deadline. A total of 90 runs were received; activity were presented and discussed in Links: runs were submitted for all tasks and for detail. Both traditional and innovative http://www.clef-campaign.org/ all topic languages. This was a very good strategies had been adopted to address the result for the first year of activity, CLIR task; various approaches had been Please contact: especially in consideration of the severe experimented for query expansion and Carol Peters Ð IEI-CNR (CLEF Coordinator) time constraints on participation. results merging. Preliminary papers Tel: +39 050 315 2897 E-mail: [email protected] describing the approaches adopted by the

ERCIM News No. 44, January 2001 63 EVENTS

CALL FOR PARTICIPATION recently renovated. Students and visualization, large data sets, surface lecturers will reside in the monastery for extraction, human factors in First DELOS International the duration of the School, giving visualization, multi-variate visualization, Summer School on Digital students ample opportunities to know information visualization, user interaction each other and to exchange ideas among techniques, multiresolution techniques. Library Technologies themselves and with the lecturers, who Proposals of case studies from all areas of Pisa, Italy, 9-13 July 2001 are expected to stay for most of the visualization are welcome. duration of the School. The School is Further information: The DELOS summer school is one of directed by Professor Maristella Agosti, http://www.cscs.ch/vissym01/ the activities of the DELOS Network of University of Padua, Italy. E-mail: [email protected] Excellence on Digital Libraries. The Further information: DELOS NoE is an initiative funded by http://www.ercim.org/delos/ CALL FOR PAPERS the 5th Framework Programme of the European Commission. The main CALL FOR PARTICIPATION Continuity in Future Computing objective of DELOS is to contribute to Systems Ð I3 Spring Days the advancement of European DL The Fifth Pacific-Asia research and to the creation of a Workshop European DL research community. Conference on Knowledge Porto, Portugal, 23-24 April 2001 DELOS aggregates active European Discovery and Data Mining — research teams in the DL field with the PAKDD’01 One of the next major developments in intention of enhancing their computing will be the embedding of collaboration, improving the Hong Kong, 16-18 April 2001 computation, information storage, effectiveness of their research, and sensing, and communication capabilities increasing the impact of their research The PAKDD-01 conference will be an within everyday objects and achievements on the interested international forum for the sharing of environments. These new systems may applications areas. research results and practical have a profound effect on the way we experiences from the many constituent interact with computers and This is the first summer school areas of KDD, including artificial information, and ultimately the way we organised in the context of the DELOS intelligence, databases, e-commerce, work and live. The design of these NoE, and it will focus on fundamental Internet computing, , artefacts is potentially very different to issues related to the multidisciplinary high performance computing, statistics traditional interactive system design. In area of Digital Libraries. The lecturers and visualization. of the school will be leading many cases the is moved researchers in the Digital Library field. Further information: from a computer screen into the ‘real Each lecture will be devoted to a http://www.csis.hku.hk/pakdd01 world’, and interaction between user presentation of the state-of-the-art on and system is of a more continuous one of the key DL topics. CALL FOR PARTICIPATION nature. We must take into account the physical and temporal aspects of these The main purpose of the school is to Joint EUROGRAPHICS — interfaces, the use of a wide range of foster research and understanding in the IEEE TCVG Symposium modalities, and adaptive technologies. fundamental technologies underlying the Digital Libraries field. It is directed on Visualization As part of the I3 (Intelligent towards members of the research Information Interfaces) Spring Days Ascona, Switzerland, community in the wide sense, that is, event, the partners of the TMR project 28-30 May 2001 graduate students and young TACIT (Theory and Applications of researchers and professionals involved Continuous Interaction Techniques), Following Vienna’99, and Amsterdam’00, in R&D in DL-related areas, possibly which includes several ERCIM member the third joint Eurographics – IEEE representing the information organisations, will hold a workshop to TCVG Symposium on Visualization technology scientist, the industrial focus on continuous interaction issues will be held in Ascona, Switzerland. communities (electronic publishing, in future computing systems, and The 3-day programme will include broadcasting, software industry, etc.) modelling to support the design of such research papers, case studies and invited and the user communities interested in systems. presentations. Students are particularly Digital Libraries technologies (libraries, encouraged to participate. Participation is by means of paper archives, museums). submission – see the link below for Topics The School will be held in Pisa in the more details. Submission deadline: 15 Suggested topics for research papers beautiful and evocative setting of the February (full papers), 2 March include: , Internet- Santa Croce in Fossabanda (position papers). based visualization, visualization (http://www.fossabanda.it), an ancient systems, volume rendering, database Further information: monastery dating from the 14th century http://kazan.cnuce.cnr.it/TACIT/I3WS.html

64 ERCIM News No. 44, January 2001 EVENTS

CALL FOR PAPERS SPONSORED BY ERCIM • Dissertation submission: 15 January 2001 International Symposium MFCS 2001 Ð • Industrial Application submission: on Virtual and Augmented 26th International Symposium 12 March 2001 • Tools request and submission: Architecture Ð VAA01 on Mathematical Foundations 12 March 2001 Dublin, 21-22 June 2001 of Computer Science • Tutorial submission: 12 February 2001 Marianske Lazne, Czech This symposium aims to bring together Republic, 27-31 August 2001 Further information researchers investigating the http://www.dsi.unifi.it/icsm2001/ intersection of architecture, computer The series of MFCS symposia, CALL FOR PAPERS graphics and computer vision. Example organized alternately in Poland, applications of this fusion include: Slovakia, and the Czech Republic since Journal of Intelligent representations of real buildings in 1972, has a long and well-established virtual environments, historical and tradition. The MFCS symposia Information Systems, Special cultural setting reconstruction and encourage high-quality research in all Issue on Automated Text presentation, virtual graphics overlay of branches of theoretical computer Categorization real buildings for navigational support, science. Their broad scope provides an education, or architectural presentation. opportunity to bring together specialists Guest Editors: Thorsten Joachim, GMD The symposium will have a mixture of who do not usually meet at specialized and Fabrizio Sebastiani CNR; Published invited lectures by established conferences. Papers presenting original by Kluwer Academic Publishers researchers and research presentations research on theoretical aspects of describing current results. Topics of computer science are sought. The Seeked are submission of high quality, interest (but not limited to) include: 3D submission deadline is 15 March 2001. original work that has not been Reconstruction, Architectural submitted, accepted for publication, or Photogrammetry, plenoptic recovery, Further information: http://math.cas.cz/~mfcs2001/ published elsewhere, covering any uncalibrated recovery of buildings from aspect of automated text categorization, video, range sensors for environment CALL FOR PAPERS including but not restricted to: scanning, multi-scan fusion, data • Machine learning methods for text representations for environments, use ICSM Ð International Conference categorization of architectural knowledge, on • Theoretical models of text media/tourism applications, categorization architectural applications, construction Florence, Italy, 6-12 November 2001 • Hierarchical text categorization industry applications, video • Text analysis and indexing methods registration, data representations for ICSM is the major international for text categorization environments, use of architectural conference in the field of software and • Dimensionality reduction for text knowledge, media/tourism applications, systems maintenance, evolution, and categorization architectural applications, construction management. The focus of the • Evaluation issues in text industry applications, video conference will be to explore the new categorization registration, interaction in virtual and challenges that the Internet, as a driver • Applications of text categorization augmented environments, user for business changes, poses for • Automated categorization of Web navigation in virtual buildings,physics software maintenance, and the new pages and Web sites in virtual buildings, realistic rendering opportunities it opens as infrastructure • Text filtering and routing of architecture, hardware for and enabling technology. ICSM 2001 • Topic detection and tracking augmented reality, interactive will bring together researchers, • Spoken text categorization buildings, fusion of video with virtual practitioners, developers and users of • OCR’ed text categorization components, autonomous avatars in tools, technology transfer experts, and Important Dates augmented/virtual buildings. project managers. The Conference will • Submission deadline: be held in conjunction with SCAM – Important Dates 28 February 2001 Source Code Analysis and • Reception of papers: 15 January 2001 • Acceptance/rejection notification: Manipulation, WESS, the seventh • Notification of acceptance: 15 March 31 May 2001 Workshop on Empirical Studies of 2001 • Submission of final copy: Software Maintenance and WSE – • Reception of final papers: 9 April 31 July 2001 Workshop on WEBsite evolution 2001 • Tentative Publication Date: Further Information December 2001 Important dates http://www.dai.ed.ac.uk/VAA01/ E-mail: [email protected]. • Research Paper submission: Further Information 15 January 2001, notification of http://mason.gmu.edu/~kersch/JIIS/ Special_Issues/TextCategory.html acceptance: 1 June 2001

ERCIM News No. 44, January 2001 65 EVENTS

SPONSORED BY ERCIM • Verification and validation methods that aim at circumventing ECOOP 2001 Ð 15th European shortcomings of existing methods in Conference on Object-Oriented respect to their industrial applicability ERCIM News is the magazine of ERCIM. Programming Published quarterly, the newsletter reports on • Formal methods based conformance, joint actions of the ERCIM partners, and aims Budapest, 18-22 June 2001 interoperability and performance to reflect the contribution made by ERCIM to testing the European Community in Information Technology. Through short articles and news The ECOOP 2001 conference invites • Case studies and project reports on items, it provides a forum for the exchange of high quality technical papers reporting formal methods related projects with information between the institutes and also research results and/or experience industrial participation (e.g. safety with the wider scientific community. ERCIM critical systems, mobile systems, News has a circulation of 7000 copies. related to the object technology. Traditionally, ECOOP 2001 hosts a two object-based distributed systems) ERCIM News online edition is available at day tutorial programme. Proposals for • Application of formal mehtods in http://www.ercim.org/publication/ERCIM_News/ high-quality tutorials in all areas of standardization and industrial forums. ERCIM News is published by ERCIM EEIG, object-orientation from academic BP 93, F-06902 Sophia-Antipolis Cedex research to industrial applications are Important Dates Tel: +33 4 9238 5010, E-mail: [email protected] solicited. As in previous years, ECOOP • submission deadline: 29 March 2001 ISSN 0926-4981 2001 will host a number of workshops • notification: 22 April 2001 Director: Bernard Larrouturou • final paper: 11 May 2001 Central Editor: addressing different areas of objet- Peter Kunz oriented technology. Workshops serve More information: E-mail: [email protected] as a forum for exchanging late breaking Tel: +33 1 3963 5040 http://www.dsse.ecs.soton.ac.uk/FMICS2001/ Local Editors: ideas and theories in an evolutionary CLRC: Martin Prime stage. They typically focus on either in CALL FOR PARTICIPATION E-mail: [email protected] depth analysis or broad-ranging Tel: +44 1235 44 6555 approaches to areas related to object- SCCG 2001 Ð Spring CRCIM: Michal Haindl E-mail: [email protected] oriented technology. Conference on Computer Tel: +420 2 6605 2350 More information: CWI: Henk Nieland Graphics http://www.sztaki.hu/conferences/ieaaie2001/ E-mail: [email protected] Tel: +31 20 592 4092 Budmerice Castle, Slovak CNR: Carol Peters CALL FOR PAPERS Republic, 25-28 April 2001 E-mail: [email protected] Tel: +39 050 315 2897 FMICS 2001 Ð 6th ERCIM This event is the oldest regular annual FORTH: Constantine Stephanidis E-mail: [email protected] Workshop on Formal Methods meeting of the computer graphics Tel: +30 81 39 17 41 for Industrial Critical Systems community in Central Europe. The spring GMD: Dietrich Stobik conference attempts to cover all E-mail: [email protected] Paris, 16-17 July 2001 interesting projects from computer Tel: +49 2241 14 2509 INRIA: Bernard Hidoine graphics and image processing in CG and E-mail: [email protected] The aim of the FMICS workshops is to the applications. The philosophy of Tel: +33 1 3963 5484 provide a forum for researchers who are SCCG is to put together top experts with SICS: Kersti Hedman interested in the development and young researchers in CG and to support E-mail: [email protected] Tel: +46 8633 1508 application of formal methods in a good communication channel for East- SRCIM: Gabriela Andrejkova industry. In particular, these workshops West exchange of prospective ideas. The E-mail: [email protected] should bring together scientists that are conference chairman is Prof. Tosiyasu L. Tel: +421 95 622 1128 Kunii, from Hosei University, Japan. SZTAKI: Erzsébet Csuhaj-Varjú active in the area of formal methods E-mail: [email protected] and interested in exchanging their Tel: +36 1 209 6990 experiences in the industrial usage of Important Dates TCD: Carol O’Sullivan these methods. They also aim at the • 11 February 2001: contribution E-mail: [email protected] promotion of research and development Tel: +353 1 60 812 20 submission deadline VTT: Pia-Maria Linden-Linna for the improvement of formal methods • 11 March 2001: review notification E-mail: [email protected] and tools for industrial applications. Tel: +358 0 456 4501 • 1 April 2001: camera-ready submission deadline Free subscription Topics include, but are not restricted to: You can subscribe to ERCIM News free of • Tools for the design and development Further information: charge by: of formal descriptions http://www.isternet.sk/sccg/ ¥ sending e-mail to your local editor • Verification and validation of ¥ contacting the ERCIM office (see address above) complex, distributed, real-time ¥ filling out the form at the ERCIM website at systems and embedded systems http://www.ercim.org/

66 ERCIM News No. 44, January 2001 IN BRIEF

CWI – Edsger W. Dijkstra (70) his whole life he remained a convinced Sciences Research Council. Older Library visited his old institute CWI in marxist, which brought him into codes are available in an HSL Archive October, 2000, where he delivered a difficulties before WWII in The Library for research purposes to anyone lecture on combating complexity in Netherlands, and afterwards in the USA, without charge. For details of both problem solving. Dijkstra, now emeritus where he was temporarily dismissed from libraries, see http://www.cse.clrc.ac.uk/ professor of the University of Texas at his post in the 1950s. He remained active Activity/HSL/. Austin, where he went to in 1984 after until the end, still handling all his positions at Eindhoven University of correspondence himself. Technology and Burroughs, graduated from Leiden University as a theoretical SZTAKI actively participated in physicist. He worked at CWI in the period the Cactus testbed demo of the 1952-1962. In 1959 he defended his PhD European Grid Forum in Dallas at the in Computer Science. He is generally Supercomputing conference 4-10 considered as the father of structured November. The Cactus testbed demo demonstrated the usage of a European Grid connecting the following sites: Albert-Einstein-Institut (Potsdam), Brno Supercomputing Centre, the DAS (The Netherlands), Konrad-Zuse-Zentrum (Berlin), SZTAKI Edsger (Budapest), Poznan Supercomputing and W. Dijkstra during his Networking Centre, Padernborn Centre visit to for . The program that Commissioner Busquin visits CWI CWI. was partially demonstrated by Ferenc Szalai (staff member of SZTAKI) was programming (one of his famous slogans solving 3D wave equations related to the in the late 1960s was ‘GO TO considered collision of black holes and was executed harmful’). Several of his fundamental in parallel on the machines of the concepts had considerable impact on the participating institutes. The program was development of programming languages. also able to move from site to site He received the ACM Turing Award in demonstrating a new kind of potential 1972. Dijkstra neither owned nor used a application of the Grid. For more computer until recently. information on the European Grid Forum, see http://www.egrid.org/ CWI – When Dirk Struik passed away on 21 October 2000 at the CLRC - The age of 106 years, his life had covered Group at the Rutherford three centuries. Struik was a Appleton Laboratory has released HSL 2000, a collection of Fortran packages for large scale scientific computation. The Library was started in 1963 and over the years it has evolved and has been extensively used on a wide range of computers. The latest release of the Library, HSL 2000, contains codes for Automatic Differentiation, Differential Equations, Eigenvalues and Eigenvectors, Mathematical Functions, Sorting, Linear Programming, Linear Algebra, Nonlinear Mathematician Dirk Struik (1894-2000) Equations, Polynomials, Optimization and addressing a crowded audience at CWI during celebrations of his 100th birthday Non-linear Data Fitting. New packages in 1994. include state-of-the-art routines for sparse linear programming, the solution of sparse mathematician born and trained in The linear systems, and optimization. While Netherlands, which he left when Norbert HSL 2000 is a commercial product, it is Wiener asked him in the 1930s to come also available without charge to anyone to Harvard. He became famous with his working in an academic institution in the History of Mathematics (1948) which was UK. This innovation is a direct result of translated into sixteen languages. During a grant from the Engineering and Physical

ERCIM News No. 44, January 2001 67 ERCIM Ð The European Research Consortium for Informatics and Mathematics is an organisation dedicated to the advancement of European research and development, in information technology and applied mathematics. Its national member institutions aim to foster collaborative work within the European research community and to increase co-operation with European industry.

Central Laboratory of the Research Councils Rutherford Appleton Laboratory, Chilton, Didcot, Oxfordshire OX11 0QX, United Kingdom Tel: +44 1235 82 1900, Fax: +44 1235 44 5385 http://www.cclrc.ac.uk/

Consiglio Nazionale delle Ricerche, IEI-CNR Via Alfieri, 1, I-56010 Pisa, Italy Tel: +39 050 315 2878, Fax: +39 050 315 2810 http://www.iei.pi.cnr.it/

Czech Research Consortium for Informatics and Mathematics FI MU, Botanicka 68a, CZ-602 00 Brno, Czech Republic Tel: +420 2 688 4669, Fax: +420 2 688 4903 http://www.utia.cas.cz/CRCIM/home.html

Centrum voor Wiskunde en Informatica Kruislaan 413, NL-1098 SJ Amsterdam, The Netherlands Tel: +31 20 592 9333, Fax: +31 20 592 4199 http://www.cwi.nl/

Foundation for Research and Technology Ð Hellas Institute of Computer Science, P.O. Box 1385, GR-71110 Heraklion, Crete, Greece Tel: +30 81 39 16 00, Fax: +30 81 39 16 01 FORTH http://www.ics.forth.gr/

GMD Ð Forschungszentrum Informationstechnik GmbH Schlo§ Birlinghoven, D-53754 Sankt Augustin, Germany Tel: +49 2241 14 0, Fax: +49 2241 14 2889 http://www.gmd.de/

Institut National de Recherche en Informatique et en Automatique B.P. 105, F-78153 Le Chesnay, France Tel: +33 1 3963 5511, Fax: +33 1 3963 5330 http://www.inria.fr/

Swedish Institute of Computer Science Box 1263, SE-164 29 Kista, Sweden Tel: +46 8 633 1500, Fax: +46 8 751 72 30 http://www.sics.se/

Swiss Association for Research in Information Technology Dept. Informatik, ETH-Zentrum, CH-8092 Zürich, Switzerland Tel: +41 1 632 72 41, Fax: +41 1 632 11 72 http://www.sarit.ch/

Slovak Research Consortium for Informatics and Mathematics Dept.of Computer Science, Comenius University, Mlynska Dolina M SK-84248 Bratislava, Slovakia Tel: +421 7 726635, Fax: +421 7 727041 http://www.srcim.sk

Magyar Tudományos Akadémia Ð Számítástechnikai és Automatizálási Kutató Intézete P.O. Box 63, H-1518 Budapest, Hungary Tel: +36 1 4665644, Fax: + 36 1 466 7503 http://www.sztaki.hu/

Trinity College, Department of Computer Science, Dublin 2, Ireland Tel: +353 1 608 1765, Fax: 353 1 677 2204 http://www.cs.tcd.ie/ERCIM

Technical Research Centre of Finland VTT Information Technology, P.O. Box 1200, FIN-02044 VTT, Finland Tel:+358 9 456 6041, Fax :+358 9 456 6027 http://www.vtt.fi/