[standards in a NUTSHELL] Frédéric Dufaux, Gary J. Sullivan, and Touradj Ebrahimi

The JPEG XR Image Coding

PEG XR is the newest image cod- baseline JPEG, but raw encoding has imaging and other flexible image inter- ing standard from the JPEG com- very high storage capacity requirements, action usage scenarios. mittee. It primarily targets the is generally camera specific, and typically JPEG XR’s architecture reflects the representation of continuous-tone lacks interoperability and published for- new requirements specific to high/ still images such as photographic mat documentation. extended dynamic range functionalities. Jimages and achieves high image quality, JPEG XR (ITU-T T.832 | ISO/IEC The traditional baseline JPEG coding on par with JPEG 2000, while requiring 29199-2) is a new image coding system format uses a bit depth of eight for each low computational resources and storage primarily targeting the representation of the three red green blue (RGB) color capacity. Moreover, it effectively address- of continuous-tone still images such as channels, resulting in 256 values per es the needs of emerging high dynamic photographic images. It is designed to or 16,777,216 color values. range imagery applications by including address the limitations of today’s for- However, more demanding applications support for a wide range of image repre- mats and to simultaneously achieve may require a bit depth of 16, providing sentation formats. high image quality while limiting com- 65,536 representable values for each putational resource and storage capacity channel or over 2.8 * 1014 color values BACKGROUND requirements. Moreover, it aims at pro- for a three-channel RGB image. viding a new set of useful image coding Additional scenarios may necessitate MOTIVATION capabilities focused around high/extend- even greater bit depths and sample rep- Fast-evolving digital imaging technolo- ed dynamic range imagery. The “XR” resentation formats. Conversely, when gies have made digital photography a part of the JPEG XR name evokes the memory or processing power is at a tremendous success with consumers and intent of the design to apply to an premium, as few as five or six bits per professionals. At the heart of this success “extended range” of applications beyond channel may be used. To fulfill these story, the baseline JPEG image coding the capabilities of the original baseline needs and others for a very broad vari- standard (ITU-T T.81 | ISO/IEC 10918-1) JPEG standard. JPEG XR is based on the ety of applications, JPEG XR has been has played a key enabling role. However, “HD Photo” technology developed by designed to include support for a wide while it has become one of the most to address the demands of range of image representation pixel for- widely used standards in the world, the consumer and professional digital pho- mats, even including support for the nearly 20-year-old JPEG technology is tography. It is expected that JPEG XR floating-point and “” formats reaching its limits and has begun to hin- will help pioneer inventive and ground- sometimes used in the most demanding der the development of innovative fea- breaking products and services in the HDR applications. tures and performance enhancements in marketplace, and will bring widespread Besides this special focus on HDR digital photography. More recently, the added value for consumers and profes- imagery applications, other major JPEG JPEG committee produced the sionals alike. XR design objectives include better com- JPEG 2000 standard (ITU-T T.800 | ISO/ pression for enhanced quality, a unified IEC 15444-1), introducing a number of OBJECTIVES system for lossless and lossy compres- novelties and new functionalities. JPEG XR is a new compression format sion, cost-effective computational per- However, it has notably higher computa- supporting high dynamic range (HDR) formance, and new progressive decod- tional resource requirements and has formats for a new generation of digital ing features for more powerful image not made a notable impact in mobile and cameras and other imaging appli- access and manipulation. embedded environments such as the dig- cations to bring a new level of success- Finally, the JPEG committee has en- ital photography market. Serious pho- ful user experience. It is designed to couraged patent holders to allow JPEG tographers have been switching to using give cost and compression benefits for XR to be implementable free of royalty “raw” image encoding for many purposes consumer, “pro-sumer,” and profes- and license fees to foster widespread to avoid the technology limitations of sional digital photography. On top of adoption of the specification and help en- that, it also has core code fea- sure that it can be implemented by the Digital Object Identifier 10.1109/MSP.2009.934187 tures that support interactive Web widest number of organizations.

1053-5888/09/$26.00©2009IEEE IEEE SIGNAL PROCESSING MAGAZINE [195] NOVEMBER 2009 [standards in a NUTSHELL] continued

ISSUING BODY, STRUCTURE OF final standard in both ITU-T (where it TARGET APPLICATIONS THE STANDARD AND SCHEDULE is known as ITU-T Recommendation JPEG XR is intended for broad use in a The Joint Photographic Experts Group T.832) in March 2009 and in ISO/IEC very wide range of han- (JPEG) is a working group that produces (as ISO/IEC 29199-2) in June 2009. dling and digital photography applica- joint standards of three major internatio- ■ Part 3: Motion JPEG XR: This part tions. Key application target areas nal standardization organizations: specifies the use of JPEG XR encod- include the following: the International Organization for ing for stored sequences of moving ■ Robust and high-fidelity image Standardization (ISO), the International images with associated timing infor- acquisition technologies, such as a Electrotechnical Commission (IEC), and mation. The Motion JPEG XR file wide range of camera applications. the International Telecommunication format is based on the ISO Base Using JPEG XR, a more accurate rep- Union Telecom Sector (ITU-T). JPEG is Media standard. It is cur- resentation of the full range of cap- universally recognized as the leading rently a committee draft (CD) in the tured image signal fidelity can be committee for compressed image for- ISO/IEC approval process, and it is retained, avoiding the bottleneck mats, and it is responsible for the popular scheduled to reach final approval sta- introduced by the older baseline JPEG, JBIG, JPEG-LS, and JPEG 2000 tus in 2010. JPEG standard and avoiding the high families of imaging standards. ■ Part 4: Conformance testing: This storage capacity and interoperability The JPEG committee began the stan- part specifies a set of tests designed to difficulties associated with camera dardization of JPEG XR technology in verify whether code streams, files, raw image formats. July 2007. The initial design proposal encoders, and decoders meet the nor- ■ High dynamic range imaging was submitted by Microsoft, based on its mative requirements specified in Part workflows, including editing soft- HD photo technology, while the subse- 2. The tests specified provide methods ware suites and high-capability quent development and future evolution to (nonexhaustively) verify whether image processing pipelines. The abil- of JPEG XR as a standard has been the encoders and decoders meet these ity of JPEG XR to retain an extended responsibility of the JPEG committee. requirements. This part has currently dynamic range with a high degree of Under the general title “Information reached the status of final committee signal fidelity can help prevent loss Technology – JPEG XR Image Coding draft (FCD) in the ISO/IEC approval of quality in end-to-end workflow System,” the JPEG XR image coding sys- process and is on target to reach final environments. tem suite of standards (ISO/IEC 29199) approval in 2010. ■ Computationally constrained sig- consists of the following current and ■ Part 5: Reference software: This nal processing environments, such planned parts: part provides reference software for as mobile and embedded applica- ■ Part 1: System architecture: This Part 2. The reference software can aid tions. As the computational de - part is a nonnormative technical adoption of standard by providing an mands associated with JPEG XR report (TR), describing an overview of example implementation that can be images are substantially lower than different parts of the specifications used as a basis for making encoder with other coding systems such as and providing some guidelines on best and decoder products and can be used JPEG 2000 that also provide high encoding and decoding practices. It is to test conformance and interopera- quality capability. now a working draft (WD), and it is bility as well as to demonstrate the The use of JPEG XR for HDR imag- expected to become a formally pub- capabilities of the associated standard. ing is illustrated in Figure 1, as well as lished TR in 2010. It includes both encoder and decoder a comparison with JPEG and “raw” ■ Part 2: Image coding specification: functionality. This part is now at FCD encoding. When using JPEG, tone This part specifies the JPEG XR image status and is scheduled to reach final mapping or other image adjustments coding format. It was approved as a approval in 2010. must properly be applied prior to encoding, as the encoding results in a substantial irreversible loss of infor- Tone JPEG JPEG Low-Quality mation. If image adjustments are Mapping Encoding Decoding Image Rendering applied to an image after encoding and decoding it using the baseline JPEG format, serious quality degradation Tone High-Quality HDR Source Raw Mapping Image Rendering will often be evident. Raw encoding effectively supports HDR, but at the cost of high storage requirements and JPEG XR JPEG XR Tone High-Quality interoperability problems. JPEG XR Encoding Decoding Mapping Image Rendering successfully overcomes these short- comings, enabling HDR while keeping [FIG1] JPEG XR for high dynamic range imaging. efficient performance.

IEEE SIGNAL PROCESSING MAGAZINE [196] NOVEMBER 2009 TECHNOLOGY of the design is support for emerging floating point representations in the HDR imagery applications. decompressed domain. FUNCTIONALITIES JPEG XR also provides an extensive The standard specifies a syntax for In this section, we discuss the technolo- set of additional functionalities, includ- representing compressed image data, gy and design features of the JPEG XR ing lossless or sup- and it specifies the associated decoding standard. Resources for further informa- port using the same signal processing process. These processes and representa- tion about the standard and its design operations, image tile segmentation tions are generic, that is, they are appli- and capabilities are listed in “Resources.” for random access and large image for- cable to a broad range of applications The JPEG XR specification enables mats, support for low-complexity com- using compressed color and gray-scale greater effective use of compressed pressed-domain image manipulations, images in communications and comput- imagery with broad and diverse appli- support for embedded thumbnail er systems and within mobile and cation requirements. JPEG XR sup- images and progressive resolution embedded devices. ports a wide range of color encoding refinement, embedded code-stream formats including monochrome, RGB, scalability for both image resolution ARCHITECTURE CMYK, and n-component encodings and fidelity, alpha plane support, and The standard specifies the syntax and using a variety of unsigned integer, bit-exact decoder results for both fixed semantics of JPEG XR coded images and fixed point, and floating-point decoded and floating point image formats. the associated decoding process that pro- numerical representations with a vari- These features are all achieved togeth- duces an output image from the coded ety of bit depths. The primary goal is to er with high compression capability image. An input image is operated on by provide a compressed format specifica- and low computational and memory an encoder that creates a JPEG XR coded tion appropriate for a wide range of resource requirements. In fact, only image. The decoder then operates on the applications while minimizing the simple integer signal processing oper- coded image, producing an output image implementation requirements for ates are required even when compress- that is either an exact or approximate encoders and decoders. A special focus ing or decompressing images that use reconstruction of the input image

RESOURCES The JPEG XR Standard and Technical Overview • ITU-T Rec. T.832 | ISO/IEC 29199-2, “Information technology – JPEG XR image coding system – Image coding specification,” [Online]. Available: http://www.itu.int/rec/T-REC-T.832. • S. Srinivasan, tag">C. Tu, S. L. Regunathan, and G. J. Sullivan, “HD Photo: A new image coding technology for digital photogra- phy,” in SPIE Appl. Dig. Image Proc. XXX, vol. 6696, paper 6696-0A, Aug. 2007. Draft Conformance Test Set and Reference Software • ISO/IEC FCD 29199-4: Information technology—JPEG XR image coding system—Part 4: Conformance testing [ISO/IEC JTC 1/SC 29/WG 1 N 5017] [Online]. Available: http://www.itscj.ipsj.or.jp/sc29/open/29view/29n10428c.htm. • ISO/IEC FCD 29199-5: Information technology—JPEG XR image coding system—Part 5: Reference software [ISO/IEC JTC 1/SC 29/WG 1 N 5020] [Online]. Available: http://www.itscj.ipsj.or.jp/sc29/open/29view/29n10430c.htm. JPEG XR High Dynamic Range Capabilities • S. Srinivasan, Z. Zhou, G. J. Sullivan, R. Rossi, S. L. Regunathan, C. Tu, and A. Roy, “Coding of high dynamic range images in JPEG XR/HD Photo,” in SPIE Appl. Dig. Image Proc. XXXI, vol. 7073, paper 7073-42, Aug. 2008. JPEG XR Signal Processing Analysis • C. Tu, S. Srinivasan, G. J. Sullivan, S. L. Regunathan, and H. S. Malvar, “Low-complexity hierarchical for lossy- to-lossless image coding in JPEG XR / HD Photo,” in SPIE Appl. Dig. Image Proc. XXXI, vol. 7073, paper 7073-12, Aug. 2008. • H. S. Malvar, G. J. Sullivan, and S. Srinivasan, “Lifting-based reversible color transformations for ,” in SPIE Appl. Dig. Image Proc. XXXI, vol. 7073, paper 7073-07, Aug. 2008. JPEG XR Performance Analysis • F. De Simone, L. Goldmann, V. Baroncini, and T. Ebrahimi, “Subjective evaluation of JPEG XR image compression,” in SPIE. Appl. Dig.Image Proc. XXXII, vol. 7443, paper 7443-21, Aug. 2009. • D. Schonberg, G. J. Sullivan, S. Sun, and Z. Zhou, “Perceptual encoding optimization for JPEG XR image coding using spatially adaptive quantization step size control,” in SPIE Appl. Dig. Image Process. XXXII, vol. 7443, paper 7443-22, Aug. 2009. Signal Processing Background Concepts • H. S. Malvar, “Fast progressive image coding without ,” in Proc. IEEE Conf., Snowbird, UT, Mar. 2000, pp. 243–252. • H. S. Malvar, Signal Processing with Lapped Transforms. Boston, MA: Artech House, 1992. • H. S. Malvar, “Biorthogonal and nonuniform lapped transforms for with reduced blocking and ringing artifacts,” IEEE Trans. Signal Processing, pp. 1043–1053, Apr. 1998.

IEEE SIGNAL PROCESSING MAGAZINE [197] NOVEMBER 2009 [standards in a NUTSHELL] continued

(depending on the fidelity represented by chrome, and arbitrary n-component are adapted dynamically based on the the encoder). The image coding specifi- color formats are supported. local statistics of coded coefficients. cation also includes a tag-based file stor- A wide range of numerical encodings Finally, the transform coefficients are age format for storage and interchange are supported at multiple bit depths. entropy encoded. For this purpose, a of such coded images. Furthermore, Formats such as 8 and 16 b, as well as variable-length coding (VLC) look-up profiles and levels, which determine con- additional specialized packed bit formats, table approach is used in which a VLC formance requirements for classes of are supported for both lossy and lossless table is selected among a small set of encoders and decoders, are specified. compression. The 32-b formats are sup- fixed predefined tables, with the table Aspects of color imagery representations ported using lossy compression. Up to 24 selection being controlled adaptively and are also addressed b of source image precision are retained based on the local statistics. in the image coding standard. through the various transforms. While Note, however, that only the decod- JPEG XR is designed such as to lead only integer arithmetic is used for inter- ing process and the coding syntax are to a cost-effective, high-performance, nal processing, lossless and lossy coding actually normatively specified in the embedded-system-friendly compression are also supported for fixed-point and standard. Figure 2 illustrates the vari- architecture. In particular, it requires a floating-point image data as well as for ous stages of the specified JPEG XR small memory footprint, and it is based integer image formats. decoding process, which is basically a on integer-only operations without using The spatial transform converts the step-by-step inversion of the expected divides. Moreover, the signal processing image data to a frequency domain rep- encoding process. structure is highly amenable to parallel resentation. A lifting-based reversible processing. Thanks to its lightweight hierarchical lapped biorthogonal trans- COMPARISON WITH processing, JPEG XR can be effectively form (LBT) is used. The transform OTHER STANDARDS implemented in low-end digital signal requires only a small number of integer The JPEG committee finalized the base- processors (DSPs) as well as in low cost processing operations for both encoding line JPEG standard (ITU-T T.81 | ISO/ application-specific integrated circuits and decoding. It is exactly invertible in IEC 10918-1) back in 1992. It has (ASICs), making it suitable for consumer integer arithmetic and hence supports become the universal predominant stan- device implementation. Furthermore, lossless image representation. The dard for digital imaging, including digi- the same signal processing operations transform is based on two basic opera- tal photography and imagery on the are used for both lossless and lossy com- tors: the core transform and the option- Internet. Its ubiquitous adoption has pression modes. al overlap filtering. The core transform made it one of the most successful inter- JPEG XR handles a wide range of is similar to the widely used discrete national standards ever established. image formats, with many of them sup- cosine transform (DCT) and can exploit However, its limited functionalities are porting HDR imagery. Supported color spatial correlation within a block- becoming an obstacle to further innova- image representations include mono- shaped region. Conversely, the overlap tion of products and services for some chrome, RGB, CMYK, and n-component. filtering is designed to exploit the corre- key application domains. The supported pixel formats cover 8- lation across block boundaries as well as The more recent JPEG 2000 standard and 16-b unsigned integer, 16- and 32-b mitigate blocking artifacts. It can be (ITU-T T.800 | ISO/IEC 15444-1) has fixed point, 16- and 32-b floating point, switched on or off by the encoder. gained market adoption in some impor- and several packed bit layout formats To enable the optimization of the tant areas, notably including digital cine- such as 5-5-5, 5-6-5, and 10-10-10 bit quantization fidelity based on the sensi- ma, geographic information systems layouts for R-G-B. tivity of the human visual system and (GIS), medical imaging, surveillance and the coefficient statistics, JPEG XR uses a identification, and the archiving of cul- TOOLS flexible coefficient quantization approach tural heritage and photographic materi- JPEG XR is based on a block transform controlled by quantization parameters als. However, JPEG 2000 has not made a design, and it uses some of the same (QPs), where the selection of a QP can notable impact in mobile and embedded high-level building blocks as in most vary in a flexible way across different environments, such as the digital pho- image compression schemes, such as spatial regions, frequency bands, and tography market. color conversion, spatial transformation, color channels. To further improve com- As JPEG XR is better tailored to the scalar quantization, coefficient scanning, pression capability, interblock coefficient needs of digital photography and partic- and entropy coding. prediction is also applied on the trans- ularly supports emerging HDR applica- The algorithm provides native sup- form coefficients to remove interblock tions and has a variety of features to port for both RGB and CMYK color redundancy. Adaptive coefficient scan- provide for flexible use while retaining types by converting these color formats ning is used to convert the two-dimen- very practical implementation character- to an internal luma-dominant format sional array of transform coefficients istics, it is expected that it will make through the use of a reversible color within a block into a one-dimensional inroads and foster innovation in the pho- transform. In addition, YUV, mono- vector to be encoded. The scan patterns tographic application domain, leading to

IEEE SIGNAL PROCESSING MAGAZINE [198] NOVEMBER 2009 Inverse- Bit Stream Entropy Coefficient Coefficient Dequantization Parsing Decoding Prediction Compressed Scanning Bit Stream

Inverse Transform

First-Level First-Level Second-Level Second-Level Inverse Core Inverse Overlap Core Inverse Overlap Color Transform Filtering Transform Filtering Transform Decoded Image

[FIG2] Basic block diagram of the normative decoding process. compelling new products enhancing Based on extensive experimentation FURTHER TECHNICAL end-user experience. by several research groups, it was DEVELOPMENTS observed that the compression perfor- The JPEG committee has initiated an TESTING AND PERFORMANCE mance of JPEG XR and JPEG 2000 exploratory activity, referred to as The JPEG committee carried out a num- implementations is typically very close advanced image coding (AIC), to study ber of experiments to thoroughly assess to each other, depending on various potential technologies for a further next the performance of JPEG XR when com- implementation details. Most of the generation of image compression systems. pared to JPEG and JPEG 2000. The cod- time, JPEG 2000 is slightly superior, Note that the term “image” should be ing schemes have been analyzed using but the difference is marginal, especial- understood in a broad sense in this con- several objective metrics and subjective ly in the range of bit rates which are text, encompassing not only still picture viewing tests. Test image data sets were meaningful for digital photography. In but also other forms of visual content selected for these experiments, represen- addition, both JPEG XR and JPEG 2000 such as image sequences, stereoscopic, tative of a variety of content relevant to consistently outperform JPEG with a multiview, volumetric, and animated vol- today’s digital photography. These high significant performance gap. umetric representations. resolution images cover a broad gamut To validate the results of the objec- Closely linked to AIC, another impor- of content with distinct characteristics. tive metrics, subjective testing has also tant activity is to define superior evalua- As JPEG, JPEG 2000 and JPEG XR been carried out. For this purpose, a tion methodologies that are appropriate are based on very different technolo- double stimulus continuous quality for the demands of current and future gies, the artifacts induced by the com- scale (DSCQS) method has been used, imagery applications. Indeed, compre- pression systems are also very different. adapted to with the evaluation of hensive and robust assessment of differ- Reliable quality assessment is therefore still images. A procedure from the Video ent imaging technologies is a critical especially challenging. Moreover, it is Quality Experts Group (VQEG) has been prerequisite step towards developing new not easy to properly weight impair- used to benchmark the results in terms image compression systems. This proves ments of different color channels to ob- of prediction accuracy, monotonicity to be a very challenging issue, due to the tain an overall assessment of the full and consistency. complexity to reliably measure image color data. Peak signal-to-noise ratio Finally, it is important to emphasis quality, either by subjective testing or (PSNR) has been one useful objective that JPEG XR is not standardizing the objective metrics. measurement for image quality mea- encoding process. As one of the attri- The JPEG Committee feels that the surement. How ever, PSNR is often butes of the design is to offer a great definition and validation of evaluation poorly correlated with subjective image flexibility for the encoder, it allows for methodologies can be useful for a wide quality. For this reason, other perceptu- future improvements and optimizations technical and scientific audience and ally driven objective metrics have also as well as competition in the market- hence foresees that this activity should been used, including single-scale and place. Therefore, considering that lead to new standard specification, for multi-scale mean structural SIMilarity JPEG XR technology has just been instance in the form of a TR. In this way, (SSIM), visual information fidelity – emerging very recently, it is expected all stakeholders in the digital imaging pixel (VIF-P), PSNR-HVS-M, and DC that future JPEG XR implementations Tune measurements. will achieve superior performance. (continued on page 204)

IEEE SIGNAL PROCESSING MAGAZINE [199] NOVEMBER 2009 best of THE WEB [ ] continued

Much work has been done and continues Two very popular graphical DSP support for converting Simulink models to be done in this area. One tool in this development packages are the Math- for use in LabView. vein is Impulse Accelerated Technologies’ works Simulink (http://www.mathworks. (http://www.impulseaccelerated.com) com/products/simulink/, http://www. CONCLUSION Impulse CoDeveloper, which can parti- mathworks.com/applications/dsp_ There are many pitfalls, some of them tion a problem between an FPGA’s comm/), and National Instruments quite subtle, that can derail a DSP proj- embedded processor and custom hard- LabView DSP (http://www.ni.com/dsp/, ect as it moves into the hardware imple- ware implemented in the FPGA, effec- ftp://ftp.ni.com/pub/devzone/pdf/ mentation stage. Since many signal tively using the FPGA logic as an tut_7006.). The Mathworks MATLAB processing experts may have little experi- accelerator for computationally expen- product is the classic command-line ence in computer engineering, we’ve sive parts of the code. design and analysis tool for DSP, com- tried to provide in this article some rec- Raising the layer of abstraction even posed of numerous toolboxes that range ommendations and Web resources to higher, graphical DSP development envi- from very generic (i.e., the signal pro- help the reader avoid as many of these ronments try to eliminate the coding cessing toolbox) to quite specific (i.e., problems as possible. requirements altogether. Here, the soft- the filter design HDL coder). Simulink ware development environment is builds on these capabilities to create a AUTHORS replaced by a drag-and-drop block dia- graphical model-based design paradigm Michael Morrow ([email protected]) gram, where the data flow is indicated by that can target hardware and software is a faculty associate in the Electrical and interconnections between functional targets. National Instruments LabView Computer Engineering Department, blocks. These environments can target provides similar capabilities but is more University of Wisconsin-Madison. hardware or software implementations. flavored toward development and imple- Cameron Wright ([email protected]) In a simple software targeted systems, mentation than analysis. National is an associate professor in the Electrical the blocks might represent subroutines Instruments also supports LabView and Computer Engineering Department, that accept data, perform a task, and directly with an extensive line of hard- University of Wyoming. return some result that is used in the ware for data acquisition and control. In Thad Welch (thadwelch@boisestate. next block. In a hardware targeted sys- a relatively recent development, LabView edu) is a professor and chair of the tem, the blocks might represent logic now has graphical blocks that will run Electrical and Computer Engineering circuits with inputs and outputs. MATLAB m-file scripts, and there is even Department, Boise State University. [SP]

[standards in a NUTSHELL] (continued from page 199) value chain could benefit from this knowl- Paper Specification (OpenXPS), which ning to emerge in a very broad variety of edge to assess their image compression specifies a set of conventions for the use other product environments. implementations as well as the associated of XML and other widely available capture and display devices. technologies to describe the content and AUTHORS appearance of paginated . Frédéric Dufaux (frederic.dufaux@epfl. PRODUCTS OpenXPS readers and printers thus ch) is a senior researcher at the Ecole JPEG XR (or its compatible HD Photo include support for JPEG XR images. Polytechnique Fédérale de Lausanne predecessor) is supported in the Microsoft Paint Shop Pro X2 includes (EPFL), Lausanne, Switzerland. He is an and operating integrated support for HD Photo / JPEG active member of the JPEG committee. systems and in some configurations of XR images, and preliminary plug-in sup- Gary J. Sullivan (garysull@microsoft. Windows XP. This includes support of port has been distributed for the Adobe com) is a video/image technology archi- various features such as image viewing in Photoshop environment (for Windows tect with Microsoft Corporation, and Windows and Mac operating systems). Redmond, Washington. He is the lead edi- Live Photo Gallery, thumbnail in Megachips Corporation of Japan has tor of the JPEG XR standard and the rap- the Windows Explorer shell, and image announced JPEG XR support in its porteur of the ITU-T visual coding experts management in the Windows Imaging embedded systems for mobile applica- group (VCEG) that manages the ITU-T Component (WIC) system. Various other tions such as camera signal processing. side of the JPEG partnership. He is an Microsoft-led efforts have included sup- Pegasus Imaging Corp. has included sup- active member of the JPEG committee. port for the feature, such as the port for the new standard in its digital Touradj Ebrahimi (touradj.ebrahimi@ SeaDragon, Photosynth, Image Composite imaging toolkit products. epfl.ch) is a professor at the Ecole Poly- Editor, and HD View initiatives. Now that JPEG XR has been approved technique Fédérale de Lausanne (EPFL), Support for JPEG XR images is a key as a final standard by ITU-T and ISO/IEC, Lausanne, Switzerland. He is an active feature of the ECMA 388 Open XML support for JPEG XR imagery is begin- member of the JPEG committee. [SP]

IEEE SIGNAL PROCESSING MAGAZINE [204] NOVEMBER 2009