Documenting Paintings with Gigapixel Photography

Journal of Imaging

Article Documenting Paintings with Gigapixel Photography

Pedro M. Cabezos-Bernal 1,* , Pablo Rodriguez-Navarro 2 and Teresa Gil-Piqueras 2

1 Departamento de Expresión Gráﬁca Arquitectónica, Universitat Politècnica de València, 46022 Valencia, Spain 2 Centro de Investigación en Arquitectura, Patrimonio y Gestión para el Desarrollo Sostenible (PEGASO), Universitat Politècnica de València, 46022 Valencia, Spain; [email protected] (P.R.-N.); [email protected] (T.G.-P.) * Correspondence: [email protected]

Abstract: Digital photographic capture of pictorial artworks with gigapixel resolution (around 1000 megapixels or greater) is a novel technique that is beginning to be used by some important international museums as a means of documentation, analysis, and dissemination of their masterpieces. This line of research is extremely interesting, not only for art curators and scholars but also for the general public. The results can be disseminated through online virtual museum displays, offering a detailed interactive visualization. These virtual visualizations allow the viewer to delve into the artwork in such a way that it is possible to zoom in and observe those details, which would be negligible to the naked eye in a real visit. Therefore, this kind of virtual visualization using gigapixel images has become an essential tool to enhance cultural heritage and to make it accessible to every- one. Since today’s professional digital cameras provide images of around 40 megapixels, obtaining gigapixel images requires some special capture and editing techniques. This article describes a series of photographic methodologies and equipment, developed by the team of researchers, that have been put into practice to achieve a very high level of detail and chromatic ﬁdelity, in the documentation Citation: Cabezos-Bernal, P.M.; and dissemination of pictorial artworks. The result of this research work consisted in the gigapixel Rodriguez-Navarro, P.; Gil-Piqueras, documentation of several masterpieces of the Museo de Bellas Artes of Valencia, one of the main T. Documenting Paintings with Gigapixel Photography. J. Imaging art galleries in Spain. The results will be disseminated through the Internet, as will be shown with 2021, 7, 156. http://doi.org/10.3390/ some examples. jimaging7080156 Keywords: gigapixel photography; ultra-high resolution; art documentation; stitching; image regis- Academic Editors: Guillaume Caron, tration; virtual musealization Olga Regina Pereira Bellon and Ilan Shimshoni

Received: 14 July 2021 1. Introduction Accepted: 18 August 2021 New techniques for the dissemination of cultural heritage through digital media Published: 21 August 2021 is one of the lines of research providing beneﬁts to society since the research results can be available to the public by means online virtual museum displays. In fact, many Publisher’s Note: MDPI stays neutral important international museums [1], are using these technologies in collaboration with with regard to jurisdictional claims in the multinational Google, which has developed its own high-resolution digital capture published maps and institutional afﬁl- system for paintings. These works are exhibited on the Arts and Culture Project website [2], iations. which contains many artworks that can be displayed with a high level of detail. They are digital reproductions with a gigapixel resolution, that is, a resolution greater than 1000 megapixels, which is 50 times greater than the image resolution provided by a conventional digital camera. Copyright: © 2021 by the authors. Gigapixel images allow documenting and analysing the paintings accurately, which is Licensee MDPI, Basel, Switzerland. very useful for curators and art scholars. Furthermore, the virtual visualizations that can This article is an open access article be generated from this type of image make the artwork accessible to anyone connected to distributed under the terms and the Internet. The viewers will be able to dive into the work, in such a way, that they would conditions of the Creative Commons appreciate many details, which would be negligible to the eye in a real visit (Figure1). Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

J. Imaging 2021, 7, 156. https://doi.org/10.3390/jimaging7080156 https://www.mdpi.com/journal/jimaging J. Imaging 2021, 7, 156 2 of 20 J. Imaging 2021, 7, x FOR PEER REVIEW 2 of 20

FigureFigure 1. 1.Hans Hans Holbein Holbein the the Younger, Younger, 1533, 1533,The AmbasssadorsThe Ambasssadors. Picture. Picture and detail and obtaineddetail obtained from Google from ArtsGoogle and Arts Culture and Culture [2]. [2].

ApartApart fromfrom Google,Google, therethere are veryvery few companies specialized in capturing gigapixel imagesimages ofof artworksartworks duedue toto thethe technicaltechnical complexity and the specialized equipment. There There areare somesome examples,examples, suchsuch asas thethe FrenchFrench statestate organization organizationCentre Centre de de Recherche Recherche et et de de Restaura- Restau- tionration des des Mus Muséesées de de France FranceCR2MF CR2MF [3 ],[3], the the Italian Italian company companyHaltadefinizione Haltadefinizione[4 ],[4], or or the the Spanish Span- Madpixelish Madpixel [5]. Some[5]. Some researchers researchers have have used used gigapixel gigapixel photography photography for documenting for documenting rupes- trianrupestrian paintings paintings [6] and [6] even and combined even combined with multispectral with multispectral imaging imaging techniques techniques [7]. Other [7]. researchOther research works haveworks used have gigapixel used gigapixel images images for documenting for documenting large Baroque large Baroque illusionistic illu- paintingssionistic paintings [8,9]. [8,9]. DigitalDigital photographicphotographic capturecapture withwith gigapixelgigapixel resolutionresolution is not an easy task and and serious serious difficultiesdifficulties may arise arise due due to to physical physical problems problems such such as the as thediffraction diffraction of light, of light, which which rep- representsresents a barrier a barrier that that limits limits the thesharpness sharpness that thatcan canbe obtained be obtained with with an optical an optical device device and anda digital a digital sensor. sensor. For this For thisreason, reason, the progression the progression in the in resolution the resolution of digital of digital sensors sensors has hasalready already reached reached the limit the limitestablished established by the by diffraction the diffraction of light of and light by andthe optical by the resolu- optical resolutiontion provided provided by the bylenses the lenses[10] that [10 cannot] that cannot match matchthe resolution the resolution of the current of the currenttop digital top digitalsensors. sensors. Going beyond Going beyond this would this wouldnot take not advantage take advantage of the effective of the effective resolution resolution of the ofsensor, the sensor, unless unless the size the of size the ofoptical the optical-sensor-sensor assembly assembly were increased, were increased, which is which not practi- is not practicalcal for the for future the future development development of conventional of conventional cameras. cameras. In fact, In fact,there there have have been been few fewat- attemptstempts to to develop develop prototype prototype cameras cameras providing providing gigapixel gigapixel images images on on the the fly. fly. One One of of the the firstfirst waswas the large format format camera camera developed developed within within the the GigapxlGigapxl Project Project byby Graham Graham Flint Flint in in2000 2000 [11 [].11 In]. this In this bulky bulky camera, camera, the theimage image was was exposed exposed on a on 450 a × 450 225× mm225 negative mm negative film, film,which which was later was digitized later digitized to form to a form4 gigapixel a 4 gigapixel (Gp) image. (Gp) Another image. Anothermore recent more proposal recent proposalis the AWARE is the 2 AWARE gigapixel 2 gigapixelcamera, started camera, by startedD. Brady by and D. his Brady team and in 2012. his team This in proto- 2012. Thistype prototypeincluded 98 included micro-cameras 98 micro-cameras with a combined with a combinedresolution resolutionof 1 Gp and of it 1 is Gp still and being it is stilldeveloped being developedand improved. and improved.The AWARE The 10 AWARE and 40 models 10 and reach 40 models almost reach 3 Gp. almost Neverthe- 3 Gp. Nevertheless,less, such special such prototypes special prototypes are far from are becoming far from becomingan accessible an accessibleoption due option to their due high to theircomplexity high complexity and cost. and cost. AA reasonablereasonable solutionsolution to overcome the problem of diffractiondiffraction and achieve gigapixel imagesimages usingusing conventionalconventional camerascameras is multi-shotmulti-shot panoramic capture, which consists in in obtainingobtaining aa setset ofof photographsphotographs from the same point of view and with a sufficient sufficient overlap overlap betweenbetween adjacentadjacent picturespictures soso that,that, by means of image s stitchingtitching software, they can be be joined joined toto composecompose aa higherhigher resolutionresolution imageimage [[12].]. InIn orderorder toto achieveachieve aa perfectperfect stitch between pictures and to avoid parallax errors, it is is mandatorymandatory to useuse a panoramicpanoramic head (Figure 22)) toto fixfix thethe positionposition ofof thethe opticaloptical centrecentre oror no-parallaxno-parallax point of the lens, while rotating the camera to obtain the different shots that will compose the final image. Moreover, a telephoto lens should be used so as to maximize the final resolution.

J. Imaging 2021, 7, x FOR PEER REVIEW 3 of 20

J. Imaging 2021, 7, 156 3 of 20 will compose the final image. Moreover, a telephoto lens should be used so as to maximize the final resolution.

Figure 2. Manual and motorized panoramic panoramic heads. heads.

TThehe capture system developed by by Google, Google, called called Art Art Camera Camera [ [1313],], seem seem to to be be based based on this this same same principle, principle, and and consists consists of ofa camera a camera integrated integrated in a inpanoramic a panoramic head headthat pro- that progressivelygressively sweeps sweeps the the painting painting from from a fixed a fixed point. point. The The partial partial images images are are subseq subsequentlyuently processed by the multinational itself. This type type of of camera camera is is not not for for sale sale and and there there are are only only a very limited number of units available exclusively to to the the company. company. The capture systems used used by by other other commercial commercial companies companies such such as as Haltadefinizione, Haltadefinizione, are notnot fullyfully described,described, butbut some some of of them them are are based based on on using using high-end high-end digital digital reflex reflex cameras cam- anderas high-costand high- automatedcost automated panoramic panoramic heads heads by Clauss by Clauss [14]. Madpixel[14]. Madpixel has his has own his automatic own au- panoramictomatic panoramic head called head MadpixelROB called MadpixelROB [15]. The [1 aforementioned5]. The aforementioned research worksresearch have works also usedhave thealso panoramic used the panoramic head methodology. head methodology. Using aa panoramicpanoramic headhead and and a a telephoto telephoto lens lens is ais verya very effective effective method method for document-for docu- ingmenting relatively relatively small small paintings paintings with with gigapixel gigapixel resolution, resolution, but but when when it comesit comes to to capturing captur- artworksing artworks of moderate of moderate size, size, there there arise arise some some drawbacks drawbacks that that limit limit image image sharpness sharpness due due to narrowto narrow depth depth of fieldof field provided provided by longby long focal focal length length lenses. lenses. This This problem problem will bewill discussed be dis- andcussed fixed and by fixed following by following the techniques the techniques that will that be will revealed be revealed below, below since, since one of one the of main the aimsmain ofaims this of article this article is to show is to ashow new a and new very and accurate very accurate gigapixel gigapixel capture capture methodology methodol- that usesogy that relatively uses relatively affordable affordable equipment. equipment. Moreover, Moreover, the digital the image digital processing image processing to generate to thegenerate gigapixel the gigapixel image and image the dissemination and the dissemination of the results of the on results the Internet on the will Internet be explained will be andexplained carried and out carried by means out ofby opensourcemeans of opensource software. software.

2. Materials and Methods 2.1. Capture Techniques 2.1. Capture Techniques The photographic equipment that was used in this research work consisted of a The photographic equipment that was used in this research work consisted of a 32.5 32.5 Mp digital camera Canon EOS 90D, equipped with a fixed telephoto lens Canon EF Mp digital camera Canon EOS 90D, equipped with a fixed telephoto lens Canon EF 200 200 mm f2.8 L II USM, a Manfrotto tripod 28B and a Manfrotto panoramic head 303 SPH. mm f2.8 L II USM, a Manfrotto tripod 28B and a Manfrotto panoramic head 303 SPH. Long focal length lenses should be used to maximize resolution and the camera must Long focal length lenses should be used to maximize resolution and the camera must be quite close to the artwork, so that, in these circumstances, depth of field provided by be quite close to the artwork, so that, in these circumstances, depth of field provided by the lens is very short. Furthermore, this is aggravated by the need to use an intermediate the lens is very short. Furthermore, this is aggravated by the need to use an intermediate diaphragm aperture to avoid losing global sharpness by the effect of light diffraction [16]. diaphragm aperture to avoid losing global sharpness by the effect of light diffraction [16]. This problem causes a loss of sharpness in certain zones of the off-centred shots of the This problem causes a loss of sharpness in certain zones of the off-centred shots of the paintings, when using the panoramic head capture methodology. This is due to the angular paintings, when using the panoramic head capture methodology. This is due to the angu- deviation between the camera sensor and the painting. lar deviation between the camera sensor and the painting. When the digital sensor remains parallel to the painting, all the points are perfectly When the digital sensor remains parallel to the painting, all the points are perfectly focused, so the distance between them and the sensor remain the same. However, as thefocused, camera so isthe rotated distance towards between the them picture and margins, the sensor the remain angle betweenthe same. the However, sensor and as the the paintingcamera is increasesrotated towards dramatically the picture (Figure margins,3). This the causes angle between only those the points sensor thatand arethe paint- at the sameing increases distance dramatically from the camera (Figure sensor 3). This as thecauses focused only pointthose topoints be perfectly that are sharp,at the same while those remaining at different distances, will be gradually blurred due to narrow depth of field (Figure4).

J. Imaging 2021, 7, x FOR PEER REVIEW 4 of 20 J. Imaging 2021, 7, x FOR PEER REVIEW 4 of 20

J. Imaging 2021 7 distance from the camera sensor as the focused point to be perfectly sharp, while those , , 156 distance from the camera sensor as the focused point to be perfectly sharp, while those4 of 20 remaining at different distances, will be gradually blurred due to narrow depth of field remaining at different distances, will be gradually blurred due to narrow depth of field (Figure 4). (Figure 4).

FigureFigure 3.3. ObliquityObliquity between the shots and the canvas when using a panoramic head head.. Figure 3. Obliquity between the shots and the canvas when using a panoramic head.

Figure 4. Loss of sharpness in an oblique shot due to narrow depth of field of long focal lenses. FigureFigure 4.4. LossLoss ofof sharpnesssharpness in an oblique shot due to narrow depth of field field of long focal lenses lenses.. To solve this problem, a new capture technique has been put into practice, which ToTo solvesolve thisthis problem,problem, aa newnew capturecapture techniquetechnique has been put into practice,practice, which consists in taking the shots while moving the camera parallel to the painting (Figure 5). consistsconsists inin takingtaking the shots while moving the the camera camera parallel parallel to to the the painting painting (Figure(Figure 5).5) . Doing so, the sharpness of the pictures will be always optimal and never limited by depth DoingDoing so,so, thethe sharpnesssharpness of of the the pictures pictures will will be be always always optimal optimal and and never never limited limited by by depth depth of of field. However, while this method would solve the problem of sharpness, it also gen- field.of field. However, However, while while this this method method would would solve solve the problemthe problem of sharpness, of sharpness, it also it generatesalso generates some difficulties that must be overcome. On the one hand, as the viewpoint would someerates difficulties some difficulties that must that bemust overcome. be overcome. On the On one the hand, one hand, as the as viewpoint the viewpoint would would vary vary continuously, the light reflected by the work will change between the shots, causing continuously,vary continuously, the light the reflectedlight reflected by the by work the willwork change will change between between the shots, the shots, causing causing slight slight differences in exposure and problems with specular reflections. Those issues can be differencesslight differences in exposure in exposure and problems and problems with specular with specular reflections. reflections. Those issuesThose canissues be can solved be by using a strategically placed controlled light source, which moves along with the camera.

J. Imaging 2021, 7, x FOR PEER REVIEW 5 of 20

J. Imaging 2021, 7, 156 5 of 20 solved by using a strategically placed controlled light source, which moves along with the camera.

FigureFigure 5.5. Parallel camera translation to take all the pictures frontally to the canvas canvas..

AnotherAnother problem arises when it comes to joiningjoining the photo mosaic mosaic,, obtained in in that that way,way, soso most of of the the current current software software for for stitching stitching images images would would be unable be unable to join to the join mo- the mosaicsaic when when the the viewpoint viewpoint is isnot not static. static. Fortunately, Fortunately, this this problem problem can can be be solved solved with with the the helphelp ofof somesome stitchingstitching algorithms,algorithms, initially developed by the German professor Helmut DerschDersch [[1177],], which which were were implemented implemented in in Hugin Hugin [18 [1], an8], opensourcean opensource stitching stitching software software with generalwith general public public license license that allows that allows joining joining pictures pictures taken fromtaken different from different viewpoints, viewpoints, as long asas theylong representas they represent a planar surface.a planar Optionally, surface. Optionally the commercial, the commercial software PTGUI software [19 ]PTGUI would also[19] dowould an excellent also do an job excellent when stitching job when these stitching kinds these of photomosaics. kinds of photomosaics. AnotherAnother wayway toto construct construct the the gigapixel gigapixel image image is is by by means means SfM SfM (structure (structure from from motion) mo- photogrammetrytion) photogrammetry software, software, since since the constructionthe construction algorithm algorithm is conceivedis conceived to to generate generate a textureda textured mesh mesh from from pictures pictures taken taken from from different different viewpoints. viewpoints. The problem is that it it wouldwould bebe neededneeded aa greatergreater numbernumber of shots since a minimum overlap of 50% between adjacentadjacent pictures is advisable, while a 30% overlap would be fine fine when using stitching software.software. However, an advantage of using SfM photogrammetry software software would would be be the the 3D3D reconstructionreconstruction of the painting and even the frame, which is not a plane surface, surface, wh whileile stitchingstitching softwaresoftware would would provide provide an an accurate accurate 2D 2D rectified rectified orthophoto orthophoto of theof the canvas canvas but but not ofnot the of volumetricthe volumetric frame. frame. TakingTaking into into account account the the previous previous considerations, considerations, the following gigapixel capture capture techniquestechniques areare proposed,proposed, whichwhich willwill bebe suitablesuitable for paintingspaintings or planar surfaces: TheThe single-viewpointsingle-viewpoint capture capture technique technique is ais known a known and common and common methodology, methodology, which consistwhich inconsist using in a panoramic using a panoramic head to rotate head the to camera rotate the around camera the no-parallax around the point, no-paral whichlax wouldpoint, which be fixed would and centredbe fixed inand front centred of the in canvas, front of thus the canvas, obtaining thus a setobtaining of overlapping a set of picturesoverlapping (Figure pictures6). This (Figure technique 6). This requires technique limiting requires the obliquity limiting ofthe the obliquity shots in of relation the shots to thein relation canvas to in the order canvas to preserve in order the to maximumpreserve the image maximum sharpness. image sharpness. TheThe parallel-multi-viewpointparallel-multi-viewpoint c captureapture t techniqueechnique is is a a new method, which consist consist in in takingtaking aa mosaicmosaic ofof overlappedoverlapped pictures as the camera is moved in in parallel to the canvas describingdescribing several rows rows or or columns. columns. The The camera camera digital digital sensor sensor must must be parallel be parallel to the to art- the artwork,work, or, or,in other in other words, words, the lens the optical lens optical axis should axis should be orthogonal be orthogonal to the canvas to the (Figure canvas (Figure7). 7).

J. Imaging 2021, 7, 156 6 of 20 J. Imaging 2021, 7, x FOR PEER REVIEW 6 of 20

J. Imaging 2021, 7, x FOR PEER REVIEW 6 of 20

FigureFigure 6.6. Single-viewpointSingle-viewpoint capturecapture technique.technique. TheThe cameracamera isis rotatedrotated aroundaround itsits no-parallaxno-parallax point.point. Figure 6. Single-viewpoint capture technique. The camera is rotated around its no-parallax point.

Figure 7. Parallel-multi-viewpoint capture technique. The camera moves parallel to the canvas.

J. Imaging 2021, 7, x FOR PEER REVIEW 7 of 20

J. Imaging 2021, 7, 156 7 of 20

Figure 7. Parallel-multi-viewpoint capture technique. The camera moves parallel to the canvas.

TheThe tilted-multi-viewpointtilted-multi-viewpoint capturecapture techniquetechnique is aa variationvariation ofof thethe previousprevious one,one, inin whichwhich thethe cameracamera cancan bebe tiltedtilted ifif required.required. ThisThis optionoption maymay seemseem awkwardawkward butbut cancan bebe usefuluseful inin thethe casecase ofof capturingcapturing largelarge paintingspaintings withwith aa tripodtripod thatthat isis notnot talltall oror shortshort enoughenough toto allowallow thethe parallelparallel movementmovement ofof thethe cameracamera inin frontfront ofof thethe wholewhole painting.painting. In thatthat case,case, thethe cameracamera couldcould bebe tiltedtilted whenwhen takingtaking thethe upperupper oror lower rows to cover the entire canvas. (Figure(Figure8 8).).

FigureFigure 8. TiltedTilted-multi-viewpoint-multi-viewpoint c captureapture technique. technique. The The camera camera is moved is moved mostly mostly parallel parallel to the to can- the vas and can be tilted if needed. Normally, only the upper or lower rows would be tilted. canvas and can be tilted if needed. Normally, only the upper or lower rows would be tilted.

TheThe resolutionresolution of of the the ﬁnal final gigapixel gigapixel image, image, achieved achieved with with those those methods, methods, will dependwill de- onpend several on several factors, factors such, assuch the as camera the camera sensor sensor resolution, resolution, the lens the focallens focal length, length and, and the distancethe distance between between the canvasthe canvas and and the camera,the camera, as will as will be discussed be discussed later. later.

2.2.2.2. Illumination SetSet WhenWhen capturing capturing paintings paintings there there are are several several approaches approaches to to illuminate illuminate the the artwork artwork properly.properly. TheThe simplestsimplest wayway isis toto useuse thethe museum’smuseum’s ownown illuminationillumination whenwhen thethe capturecapture processprocess isis carriedcarried outout inin situ.situ. Nevertheless,Nevertheless, forfor aa uniformuniform andand colour-accuratecolour-accurate resultresult itit wouldwould bebe advisableadvisable toto useuse dedicateddedicated andand controlledcontrolled lightlight sources.sources. InIn thisthis research,research, therethere werewere usedused twotwo MettleMettle 2828 inchinch softsoft boxesboxes lights,lights, providingproviding aa colourcolour temperaturetemperature ofof 55005500 KK andand aa CRICRI (Colour(Colour RenderingRendering Index)Index) greatergreater thanthan 90.90. TheThe softsoft boxesboxes mustmust bebe placed strategically toto avoidavoid specularspecular reﬂectionsreflections onon thethe canvascanvas andand to guarantee lighting lighting uniformity, uniformity, so so the the lighting lighting scheme scheme proposed proposed in inFigure Figure 9 is9 ad-is advisable.visable. The The lights lights must must be placed be placed outside outside the family the family of angles of zone, angles as zone, otherwise as otherwise specular specularreflections reﬂections may occur. may occur.

J. Imaging 2021, 7, x FOR PEER REVIEW 8 of 20 J. Imaging 2021, 7, 156 8 of 20 J. Imaging 2021, 7, x FOR PEER REVIEW 8 of 20

Figure 9. Lighting scheme to illuminate the artwork uniformly. In order to avoid specular reflec- FigureFiguretions on 9.9. the Lighting canvas scheme, the light toto illuminateilluminatesources must thethe be artworkartwork placed uniformly.uniformly. outside of In Inthe order order volume to to avoid avoid defined specular specular by the reﬂections reflec-family ontionsof theangles. on canvas, the canvas the light, the sources light sources must be must placed be placed outside outside of the volumeof the volume deﬁned defined by the familyby the offamily angles. of angles. WhenWhen using using the the multi-viewpointmulti-viewpoint capturecapture technique,technique, it is mandatory to move the lightinglightingWhen systemsystem using alongalong the withwithmulti thethe-viewpoint camera,camera, sosocapture aa special technique, mount was it is designed mandatory to attach to move the the soft softthe boxeslightingboxes toto system thethe tripodtripod along (Figure(Figure with 10the10).). camera, Additionally, so a special a sliding mount system was fordesigned the tripod’s to attach feet, the which soft consistedboxesconsisted to the inin a tripoda two-wheeled two- wheeled(Figure aluminium10). aluminium Additionally, bar, bar, was wasa alsosliding also constructed constructedsystem for to the facilitateto tripod’sfacilitate the feet, movementthe whichmove- ofconsistedment the of set. the in set.a two -wheeled aluminium bar, was also constructed to facilitate the movement of the set.

FigureFigure 10.10. LightingLighting mountmount to attach thethe lighting system to the tripod. Figure 10. Lighting mount to attach the lighting system to the tripod. 2.3.2.3. CameraCamera SettingsSettings and Colour Accuracy 2.3. CameraItIt isis advisableadvisable Settings to toand use use Colour an an intermediate intermediate Accuracy diaphragm diaphragm aperture aperture value value to to avoid avoid the the diffraction diffrac- phenomenontion phenomenonIt is advisable while whileto maintaining use maintaining an intermediate good good depth diaphragm depth of ﬁeld. of field. aperture Setting Setting thevalue the camera tocamera avoid to f8 tothe wouldf8 diffrac- would be advisabletion phenomenon as it is usually while themaintaining golden number good depth for maximizing of field. Setting image the sharpness. camera to The f8 camerawould speed will depend on the lighting conditions. Slow speeds will not be a problem since a tripod is used.

J. Imaging 2021, 7, x FOR PEER REVIEW 9 of 20

J. Imaging 2021, 7, 156 9 of 20 be advisable as it is usually the golden number for maximizing image sharpness. The camera speed will depend on the lighting conditions. Slow speeds will not be a problem since a tripod is used. TheThe cameracamera ISOISO valuevalue mustmust be atat thethe lowerlower settingsetting to avoid noisy pictures. An ISO 100100 valuevalue isis advisable.advisable. RAW ImageImage filefile formatformat should be used instead of JPG in other to maximizemaximize thethe dynamicdynamic range and to properly adjust the white balance and to apply an accurateaccurate colourcolour profileprofile duringduring thethe digitaldigital developingdeveloping process. WhenWhen usingusing telephototelephoto lenses,lenses, it isis advisable to shoot the the camera camera remotely, remotely, even when when aa tripodtripod isis used.used. For this reason, it is is recommended recommended using using a a remote remote shooter shooter or, or, even even better, better, aa mobilemobile devicedevice oror notepadnotepad withwith aa compatiblecompatible application for shooting the camera. In this research,research, aa notebooknotebook withwith thethe freefree softwaresoftware EOSEOS Utility by Canon, was used to control the cameracamera viavia Wi-Fi.Wi-Fi. InIn orderorder toto capturecapture thethe colourscolours ofof the artwork accurately, a Xrite ColorCheckerColorChecker chart waswas usedused to create a a colour colour profile profile from from a apicture picture of ofthe the chart chart,, which which was was illumin illuminatedated by bythe the scene scene lights lights (Figure (Figure 11). The11). colour The colour profile profile was generated was generated by the Color by theChecker ColorChecker Camera CameraCalibration Calibration software softwareby Xrite and by Xrite it can and be used it can with be used RAW with developers RAW developers,, such as the such open- as thesource opensource RawTherap RawTherapeeee [20], as well [20 as], with as well other as commercial with other commercialsoftware, such software, as Adobe such Cam- as Adobeera Raw Camera or Adobe Raw Lightroom or Adobe [ Lightroom21]. The picture [21]. Theof the picture chart ofcan the also chart be used can also as a be reference used as ato reference adjust the to proper adjust thewhite proper balance white of all balance the shots of all at theonce, shots by atusing once, the by middle using thegrey middle patch greyof the patch chart of to the neutralize chart to any neutralize colour anypredominance. colour predominance.

FigureFigure 11.11.Reference Reference picture picture with with the the X-Rite X-Rite ColorChecker ColorChecker chart chart that that will will help help to set to the set proper the proper white balancewhite balance and to and create to acreate speciﬁc a specific colour colour proﬁle profile for the for lighting the lighting conditions. conditions.

2.4.2.4. GeneratingGenerating thethe GigapixelGigapixel Image InIn orderorder to to join join the the set set of picturesof pictures obtained obtained with with any any of the of aforementioned the aforementioned techniques, tech- aniques, stitching a stitching software software such as Huginsuch as or Hugin PTGUI or can PTGUI be used. can be As used. mentioned As mentioned before, Hugin before, is aHugin powerful is a opensourcepowerful opensource stitching software stitching that software can even that calculate can even the calculate camera the translation camera betweentranslation shots, between when theshots, multi-viewpoint when the multi technique-viewpoint is used. technique The only is used. drawback The only of Hugin draw- is thatback is of slow Hugin when is that compared is slow towhen the commercialcompared to alternative the commercial PTGUI, alternative which is incrediblyPTGUI, which fast. is incrediblyThe workflow fast. with the stitching software is very simple. Initially, the set of pictures are analysedThe workflow to automatically with the stitching detect homologoussoftware is very points simple. in the Initially, overlapping the set zones of pictures of the shots.are analysed Then, the to auto optimizermatically algorithm detect useshomologous all these points points in to the calculate overlapping the spatial zones position of the of every picture, which is defined by the pitch, yaw, and roll angles. In addition to this, the camera translations, and the parameters to correct the radial distortion, caused by the lens, are also computed. The final step allows rendering the set of pictures into a whole image at

J. Imaging 2021, 7, 156 10 of 20

gigapixel resolution. The maximum final resolution will depend on the number of shots and their original resolution. That is estimated by the software so as to avoid incremental interpolations. When using the multi-viewpoint technique, it is possible to assemble the gigapixel image with an automated SfM photogrammetry software, such as Agisoft Metashape. As mentioned before, using this kind of software would require a greater overlapping between adjacent shots and it is much more time consuming. This option would be interesting only when documenting volumetric artworks, such as altar pieces. The workflow for this kind of software begins with the determination of the camera positions from the analysis of the homologous points and the sparse point cloud. Then, the dense cloud of the model can be computed, and can be later triangulated to form a polygonal mesh. This mesh can be texturized by projecting the original pictures onto the mesh. Finally, an orthophoto of the 3D textured model can be generated at gigapixel resolution. It should be mentioned that SfM photogrammetric techniques would not provide an absolutely accurate three-dimensional restitution of those subtle surface details, but since our main objective is obtaining an orthophoto of the painting, it would provide excellent results. In order to render fine surface properties of the painting, reflectance transformation imaging (RTI) would be very useful, as well as other much more expensive 3D scanning techniques [22].

2.5. Canvas Measurement Knowing the real dimension of the paintings is important to provide an accurate scale of the gigapixel image. Normally the dimensions of the artwork are specified on the wall labels, but they are approximated. A Topcon Image IS robotic total station was used for taking the measurement of the canvases without touching the artwork. This total station has a reflector-less measurement capability between 1.5 and 250 m, providing a short-range precision of ±5 mm (MSE), and an angular precision of 0.3 mgon. The instrument was placed in front of the painting, and, with the help of the telescopic sight, the corners of the canvas were identified. The coordinates of these points were obtained using TopSurv, a software that comes with the total station.

3. Results In this section, some examples, which were captured using the described techniques, will be exposed. All the paintings belong to the Museo de Bellas Artes of Valencia (Spain).

3.1. Single-Viewpoint Capture The ﬁrst example is focused on the artwork entitled La Santa Cena, painted by Juan de Juanes in 1534. This artwork measures 92.3 cm tall × 85.8 cm wide. Due to its reduced size, it was shot using the Single-Viewpoint Technique. The camera was placed in a central position with respect to the canvas and at a distance of 2 m. In this manner, the perimeter parts were not too oblique to the camera sensor, thus preserving the image sharpness (Figure 12). All the zones had to be inside depth of ﬁeld limits provided by the telephoto lens. The diaphragm was set to f8, the ISO value to 100, and the shutter speed was set to 2 s in accordance with the lighting conditions. J. Imaging 2021, 7, x FOR PEER REVIEW 11 of 20

A reference picture, with the X-Rite Colour Checker Chart, was taken to generate a specific colour profile for the lighting conditions of the scene, and to adjust the white balance in the RAW development process precisely. In this case, only the lights of the museum were used as the lighting source. The development process was carried out using the opensource software RAW Ther- apy. There were generated 90 files in TIF format with 32 bits per channel, which were perfectly balanced with the colour profile created by the ColorChecker Camera Calibra- tion software, which is provided by the manufacturer of the colour chart. After the raw development process, the mosaic was joined using Hugin, a GPL-licensed stitching software, which generated the final image (Figure 13b), with a resolution of 26,511 × 28,520 px, that is, 756 megapixels (0.76 gigapixels). It is interesting to express the resolution relative to a measurement unit, so this value J. Imaging 2021, 7, 156 would be independent of the artwork size. This parameter is known as pixel density11 and of 20 will allow an objective comparison of the level of detail, acquired in each case. The pixel density, expressed in pixel per inch, was 787 ppi in this case.

Figure 12. Single-viewpoint capture technique. Viewpoint position in the capture of the Santa Cena Figure 12. Single-viewpoint capture technique. Viewpoint position in the capture of the Santa Cena by Juan de Juanes (1534). Museo de Bellas Artes de Valencia (Spain). by Juan de Juanes (1534). Museo de Bellas Artes de Valencia (Spain).

A total of 90 photographs were taken by rotating the camera around its centre of perspective, in steps of 2.5 degrees (horizontal rotation), and 4 degrees (vertical rotation). J. Imaging 2021, 7, x FOR PEER REVIEWThat resulted in a mosaic of 10 columns and 9 rows, with an overlapping between adjacent12 of 20

pictures greater than 30% (Figure 13a).

FigureFigure 13.13. ((aa)) SetSet ofof 9090 picturespictures takentaken duringduring thethe gigapixelgigapixel capture.capture. (b) Stitched gigapixelgigapixel Image.

AFigure reference 14 shows picture, a fragment with the ofX-Rite the resulting Colour image Checker in which Chart, the was level taken of detail to generate can be anoticed. speciﬁc The colour pixel proﬁle density for is the not lighting huge because conditions of the of long the scene, distance and between to adjust the the camera white balanceand the incanvas the RAW. This distance development was increased process precisely.to prevent In the this loss case, of sharpness only the in lights the decen- of the museumtred shots were due usedto the as obliquity the lighting between source. the canvas and the camera sensor. Nevertheless, a greater resolution could have been achieved if the proposed parallel multi-viewpoint capture method had been used, so the camera would have been placed almost at the minimum focusing distance of the lens (1.5 m instead of 2 m).

Figure 14. Fragment of the resulting gigapixel image.

3.2. Parallel Multi-Viewpoint Capture

J. Imaging 2021, 7, x FOR PEER REVIEW 12 of 20

J. Imaging 2021, 7, 156 12 of 20

The development process was carried out using the opensource software RAW Ther- apy. There were generated 90 files in TIF format with 32 bits per channel, which were perfectly balanced with the colour profile created by the ColorChecker Camera Calibration software, which is provided by the manufacturer of the colour chart. After the raw development process, the mosaic was joined using Hugin, a GPL- licensed stitching software, which generated the final image (Figure 13b), with a resolution of 26,511 × 28,520 px, that is, 756 megapixels (0.76 gigapixels). It is interesting to express the resolution relative to a measurement unit, so this value would be independent of the artwork size. This parameter is known as pixel density and willFigure allow 13. ( ana) Set objective of 90 pictures comparison taken during of the the level gigapixel of detail, capture. acquired (b) Stitched in each gigapixel case. TheImage. pixel density, expressed in pixel per inch, was 787 ppi in this case. FigureFigure 1414 showsshows aa fragmentfragment ofof thethe resulting image in which the level of detail can be be noticed.noticed. TheThe pixelpixel density density is is not not huge huge because because of of the the long long distance distance between between the camerathe camera and theand canvas. the canvas This. This distance distance was was increased increased to prevent to prevent the lossthe loss of sharpness of sharpness in the in the decentred decen- shotstred shots due todue the to obliquitythe obliquity between between the the canvas canvas and and the the camera camera sensor. sensor. Nevertheless, Nevertheless, a greatera greater resolution resolution could could have have been been achieved achievedif ifthe the proposedproposed parallelparallel multi-viewpointmulti-viewpoint capturecapture method had had been been used used,, so so the the camera camera would would have have been been placed placed almost almost at the at min- the minimumimum focusing focusing distance distance of the of thelens lens (1.5 (1.5m instead m instead of 2 ofm). 2 m).

FigureFigure 14.14. FragmentFragment ofof thethe resulting gigapixel image.image.

3.2.3.2. ParallelParallel Multi-ViewpointMulti-Viewpoint Capture The parallel multi-viewpoint capture method will produce better image quality than the previous one since the camera sensor is always parallel to the canvas. Consequently, all the zones of the pictures would be perfectly focused. Nevertheless, this approach is more complicated, so the camera must be moved along with the lights every shot, but it is worth it. This technique was put into practice in the next example to capture the painting entitled Virgen de la Leche, painted by Bartolomé Bermejo around 1478. The canvas size is quite small, measuring 49.1 cm tall × 34.5 cm wide. In this case, the previously described lighting system was used, which consisted of two soft boxed attached to the tripod (Figure 10). The lamps of the museum were turned-off so as not to mix different light sources. J. Imaging 2021, 7, x FOR PEER REVIEW 13 of 20

The parallel multi-viewpoint capture method will produce better image quality than the previous one since the camera sensor is always parallel to the canvas. Consequently, all the zones of the pictures would be perfectly focused. Nevertheless, this approach is more complicated, so the camera must be moved along with the lights every shot, but it is worth it. This technique was put into practice in the next example to capture the painting entitled Virgen de la Leche, painted by Bartolomé Bermejo around 1478. The canvas size is J. Imaging 2021, 7, 156 quite small, measuring 49.1 cm tall × 34.5 cm wide. 13 of 20 In this case, the previously described lighting system was used, which consisted of two soft boxed attached to the tripod (Figure 10). The lamps of the museum were turned- off so as not to mix different light sources. TheThe photographicphotographic set mustmust be stabilizedstabilized after changing the camera position, so the ﬂexibilityflexibility ofof thethe lightlight stand,stand, whichwhich is attached to the tripod, can leadlead to slightly blurred shotsshots duedue toto smallsmall vibrationsvibrations transmittedtransmitted to the camera. This is controlled by monitoring thethe cameracamera withwith thethe laptop.laptop. The live view image, which which is is received received via via Wi Wi-Fi-Fi from the camera,camera, is magnified magniﬁed 10 10×× to todetect detect any any slight slight vibration vibration and andto focus to focus the canvas the canvas very accu- very accurately.rately. Once Once the image the image is completely is completely steady steady and focused and focused,, the camera the camera is remotely is remotely shot (Fig- shot (Figureure 15). 15 ).

FigureFigure 15.15. RemoteRemote controlcontrol ofof the camera with the laptop via Wi-Fi.Wi-Fi.

WeWe tooktook 2020 shotsshots toto covercover thethe entireentire canvascanvas obtaining a mosaic of 5 columns and 4 4 rowsrows withwith anan overlappingoverlapping greatergreater thanthan 30%30% (Figure(Figure 1616a). The camera was placed at 11.5.5 m fromfrom thethe canvascanvas andand movedmoved horizontallyhorizontally to cover the upper row. Then,Then, the camera was movedmoved downwards vertically vertically to to initiate initiate the the second second row row and and so on. so on.A measuring A measuring tape tapewas wasattached attached to the to floor, the ﬂoor, parallel parallel to the to canvas, the canvas, to serve to serveas a reference as a reference for the fortranslation the translation move- movement.ment. The diaphragm The diaphragm was set was to setf8, the to f8, ISO the value ISO valueto 100, to and 100, the and shutter the shutter speed speedwas set was to set1/20 to of 1/20 second of second according according to the tolighting the lighting conditions. conditions.

J. Imaging 2021, 7, 156 14 of 20 J. Imaging 2021, 7, x FOR PEER REVIEW 14 of 20 J. Imaging 2021, 7, x FOR PEER REVIEW 14 of 20

Figure 16. (a) Set of 20 pictures taken during the gigapixel capture. (b) Stitched Gigapixel Image. FigureFigure 16.16. ((aa)) SetSet ofof 2020 picturespictures takentaken duringduring the gigapixel capture. ( b) Stitched Gigapixel Image. Image.

TheThe rawraw developingdeveloping process process was was carried carried out out in in the the same same way way as as the the previous previous example. exam- Nevertheless,ple. Nevertheless, the shots the wereshots stitchedwere stitched using PTGUI using insteadPTGUI ofinstead Hugin of as Hugin it is much as it faster. is much The resultingfaster. The image result (Figureing image 16b) (Figure has a ﬁnal 16b) resolution has a final of resolution 14,183 × of20,199 14,183 px, × which20,199 providespx, which a pixelprovides density a pixel of 1044 density ppi. of This 1044 is ppi. much This better is much than inbetter the previousthan in the case, previous since thecase, camera since couldthe camera be placed could closer be placed to the closer canvas. to the As canvas. mentioned, As mentioned there is no, there problem is no withproblem depth with of ﬁelddepth when of field using when this method.using this Figure method. 17 shows Figure a 17 cropped shows fragment a croppe ofd thefragment gigapixel of the image gi- centredgapixel onimage the eyecentred of the on Virgin. the eye of the Virgin.

Figure 17. Fragment of the resulting gigapixel image. FigureFigure 17.17. FragmentFragment ofof thethe resulting gigapixel image.image.

3.3. Tilted Multi-Viewpoint Capture

J. Imaging 2021, 7, x FOR PEER REVIEW 15 of 20

The last example shows the gigapixel capture of the artwork entitled Martirio de San Bartolomé, painted by Luca Giordano around 1650. The canvas size is 136 cm tall × 95 cm J. Imaging 2021, 7, 156 wide, which is a bit large to use the single-viewpoint method properly, so image sharp-15 of 20 ness would be poor in some zones of the boundary shots. Moreover, the upper and lower borders of the painting exceeded the maximum and minimum height of the tripod, so the tilted multi-viewpoint capture method was chosen in this case. 3.3. Tilted Multi-Viewpoint Capture We took 90 shots, forming a mosaic of 10 columns and 9 rows (Figure 18a). Most of the picturesThe last where example parallel shows to the canvas. gigapixel Only capture the shots of the corresponding artwork entitled to the Martirio upper and de Sanlower Bartolom rows wereé, painted slightly by tilted Luca vertically Giordano to around cover the 1650. entire The art canvaswork, size which is 136 is negligible cm tall × 95for cm the wide, image which quality. isa The bit largedistance to usefrom the the single-viewpoint camera to the canvas method was properly, set to 1. so9 m image. The sharpnesscamera settings would were be poor: diaphragm in some f8, zones ISO of100, the and boundary camera speed shots. 1/5 Moreover, of second, the accordingly upper and lowerto the borderslighting ofconditions. the painting exceeded the maximum and minimum height of the tripod, so theAs tilted in the multi-viewpoint previous cases, capture the RAW method developing was chosen process in thiswas case. carried out with RAW- TherapWeee took to apply 90 shots, the formingproper white a mosaic balance of 10 and columns the specific and 9 colour rows (Figureprofile to18 a).all Mostthe pic- of thetures pictures at once. where The stitching parallel towas the carried canvas. out Only with the PTGUI, shots correspondingwhich provided to a the gigapixel upper andim- lowerage (Fig rowsure were18b) with slightly a total tilted resolution vertically of to30 cover,884 × the44,228 entire px (1 artwork,.36 gigapixels). which isIn negligible this case, forthe thepixel image density quality. was 826 The ppi, distance something from lower the camera than in to the the previous canvas wascase setdue to to 1.9 the m. larger The cameradistance settings between were: the diaphragmcamera and f8, the ISO canvas. 100, and Figure camera 19 shows speed a 1/5 cropped of second, fragment accordingly of the togigapix the lightingel image conditions. focused on the saint’s mouth.

FigureFigure 18.18. ((aa)) SetSet ofof 9090 picturespictures takentaken duringduring thethe gigapixelgigapixel capture.capture. (b) Stitched gigapixelgigapixel i image.mage.

As in the previous cases, the RAW developing process was carried out with RAWTher- apee to apply the proper white balance and the speciﬁc colour proﬁle to all the pictures at once. The stitching was carried out with PTGUI, which provided a gigapixel image (Figure 18b) with a total resolution of 30,884 × 44,228 px (1.36 gigapixels). In this case, the pixel density was 826 ppi, something lower than in the previous case due to the larger distance between the camera and the canvas. Figure 19 shows a cropped fragment of the gigapixel image focused on the saint’s mouth.

J. Imaging 2021, 7, 156 16 of 20 J. Imaging 2021, 7, x FOR PEER REVIEW 16 of 20

FigureFigure 19.19. FragmentFragment ofof thethe resulting gigapixel image.image.

4.4. DiscussionDiscussion TheThe newnew proposedproposed multi-viewpointmulti-viewpoint gigapixelgigapixel capturecapture techniques,techniques, compared to the conventionalconventional s single-viewpointingle-viewpoint c captureapture technique technique,, improve improve the the image image quality quality of the of result- the re- sultinging gigapixel gigapixel image image.. The Theparallel parallel multi multi-viewpoint-viewpoint gigapixel gigapixel capture capture technique technique is, in is,gen- in general,eral, most most recommended, recommended, so sothere there is isno no loss loss of of sharpness sharpness produced produced by by narrow narrow de depthpth of of fieldfield providedprovided byby telephototelephoto lenses.lenses. The novel multi-viewpointmulti-viewpoint c captureapture t techniquesechniques imply usingusing differentdifferent shootingshooting methodsmethods and moving the light system along with the camera to avoidavoid changingchanging specularspecular reflectionsreflections duringduring thethe capturecapture process.process. Nevertheless,Nevertheless, the single-viewpointsingle-viewpoint c captureapture t techniqueechnique can can be also a proper method whenwhen thethe depth of of field field provides provides an an acceptable acceptable sharpness sharpness in inthe the whole whole picture picture.. The Theac- acceptableceptable sharpness sharpness would would depend depend on on the the focal focal lens, lens, the the diaphragm diaphragm aper apertureture value, value, the focusfocus distance, and and the the camera camera sensor sensor size. size. It can It canbe calculated be calculated by using by usinga free aDOF free calcu- DOF calculatorlator such suchas DOFMASTER as DOFMASTER [23]. [ 23]. FromFrom thethe experienceexperience acquired in this research research,, there are some practical practical considerations considerations thatthat cancan bebe veryvery usefuluseful forfor futurefuture researchers,researchers, whichwhich willwill bebe discusseddiscussed here. WhenWhen documentingdocumenting artworks with gigapixel resolution resolution,, t thehe pixel pixel density density is is an an im- im- portantportant aspectaspect toto bearbear inin mindmind whenwhen planningplanning thethe photographicphotographic session. A pixel density rangerange betweenbetween 600600 ppippi andand aboutabout 1000 ppi was established to guarantee an excellent level level ofof detaildetail ofof thethe paintings.paintings. The maximum reachable pixel density value is always imposed byby thethe photographicphotographic equipment, equipment, so so it dependsit depends on on the the lens lens focal focal length, length the, camerathe camera distance, dis- andtance, the and camera the camera sensor sensor resolution resol andution size. and size. InIn orderorder toto maximizemaximize thethe pixel density,density, itit wouldwould bebe advisable to use long focal lenses lenses andand toto placeplace the the camera camera as as close close to to the the painting painting as possibleas possible (the (the minimum minimum focusing focusing distance dis- oftance the lensof the can lens be can the be most the limitingmost limiting factor). factor). TheThe pixelpixel densitydensity of of the the final final gigapixel gigapixel image image can can be be calculated calculated easily easily by by Equation Equation (1), which(1), which can becan deduced be deduced geometrically: geometrically:

푆S푟r ∙·퐹퐿FL∙·푈U푓 f PD푃퐷 == ((1)1) 퐷D∙·푆S푤w where,where, 푃퐷 =PD Pixel= Pixel Density Density of the of Image the Image (ppc (ppcor ppi) or ppi)

J. Imaging 2021, 7, 156 17 of 20

Sr = Sensor Width Resolution (px) FL = Lens Focal Length (mm) Uf = Unit Conversion Factor [use 10 for obtaining the pixel density in pixels per centimetre (ppc), and 25.4 for pixels per inch (ppi)] D = Camera Distance (mm) Sw = Sensor Width (mm) Obtaining a high pixel density is not necessarily a synonym of image quality, so a picture with a great pixel density can be totally blurred. For this reason, it is crucial to choose a lens providing excellent optical quality and to use it properly. When using the proposed parallel multi-viewpoint technique, there is no problem with depth of field, so the camera sensor must remain parallel to the canvas always. The goal, in this case, is focusing the camera on the canvas as accurately as possible, maintaining the camera steady when shooting, and minimizing the loss of sharpness caused by the lens diffraction. In order to do so, it is necessary to use a tripod to maintain the camera totally stabilized. A measuring tape can be placed on the floor, parallel to the canvas, to work as a guide for the tripod legs so as to maintain the same distance from the camera to the canvas. Additionally, it is advisable to focus and shoot the camera remotely to avoid small movements that would produce slightly blurred images. This can be done by using a computer or a mobile device with the software provided by the camera. The previous aspects are critical, especially when using long focal lenses in which any small movement would be extremely amplified on the camera sensor. Moreover, setting the ISO setting to its minimum is mandatory to improve the image quality, as well as setting the proper exposure values to capture the entire dynamic range of the scene. In this sense, it is advisable to set the aperture to medium values, such as f8. Intermediate apertures would maximize the image sharpness, while avoiding diffraction problems. Once the aperture is set, the shutter speed can be determined accordingly to the diaphragm aperture to obtain the proper exposure. Visualizing the live histogram at this point can be very helpful to avoid clipping areas. The image mosaic can be planned before starting the capture to know both the horizontal and the vertical camera displacements, as well as the total amount of shots that will be necessary to cover the entire painting. As mentioned before, a minimum overlap of about 30% between adjacent shots is advisable for a successful stitching. It is also recommended to shoot the camera in portrait mode to reduce the vertical camera movements, which are less convenient than the horizontal ones. The camera vertical displacement can be determined by Equation (2):

D · S · (100 − P) C = w (2) v 100 · FL where, Cv = Camera Vertical Displacement (mm) D = Camera Distance (mm) Sw = Sensor Width (mm) P = Image Overlap Percentage (%) (a minimum of 30% is advisable) FL = Lens Focal Length (mm) The camera horizontal displacement can be easily calculated as a fraction of the vertical displacement, so the sensor aspect ratio is known:

Cv Ch = (3) Sa where, Ch = Camera Horizontal Displacement (mm) Cv = Camera Vertical Displacement (mm) Sa = Sensor Aspect Ratio (normally 3:2 for professional cameras) J. Imaging 2021, 7, 156 18 of 20

The total amount of shots composing the Photographic Mosaic (Rows and Columns) can be predicted as follows: H R = (4) Cv W C = (5) Ch where, R = Number of Rows of the Photographic Mosaic C = Number of Columns of the Photographic Mosaic H = Painting Height (mm) W = Painting Width (mm) Ch = Camera Horizontal Displacement (mm) Cv = Camera Vertical Displacement (mm) A prediction of the resolution of the final gigapixel image in Gigapixels can be obtained from the next equation: H · W · PD2 RS = (6) 1 · 109 where, RS = Total Resolution of the Gigapixel Image (Gp) H = Painting Height (cm) W = Painting Width (cm) PD = Pixel Density of the Image (ppc) As explained before, the resulting mosaic could be assembled with free stitching software, such as Hugin or with commercial one, such as PTGUI, which is very straightforward and fast. The resulting gigapixel image can be very heavy and difficult to handle unless a powerful computer is used. Fortunately, the dissemination of such heavy gigapixel images through the Internet can be carried out by decomposing the image in a multiresolution mosaic, which is also called pyramidal image. The process consists in applying an algorithm that generates different tiles from the gigapixel image, with different sizes and levels of resolution. Then, these image tiles can be loaded progressively in real time by using a specific html5 viewer. In this way, when the user demands more definition by zooming up, the viewer loads only the specific tiles for the visualization area. When visualizing the resulting multiresolution mosaic through the Internet, there is no need to use powerful equipment, since even a conventional smartphone can manage this kind of pyramidal image. This technique is used in very well-known applications such as Google Maps or Google Earth. It is important to mention that, due to the different kinds of display that the spectators can use for the visualization of the images online, it is advisable to generate the multiresolution mosaic using the standard colour profile sRGB, which is suitable for most of the devices. Nevertheless, the accuracy of the colour reproduction will depend on the gamut of the display and its proper calibration. Zoomify [23] is one of the software packages that can provide multiresolution mosaics and has its own viewers, including a free version. There are another powerful opensource viewers, such as OpenSeadragon [24], which is the one used for this project. As a result, the gigapixel images that have been shown in this article can be visualized online by using the QR codes on Figure 20 or by following the links at the caption. J. Imaging 2021, 7, 156 19 of 20 J. Imaging 2021, 7, x FOR PEER REVIEW 19 of 20

FigureFigure 20.20. ForFor anan interactiveinteractive gigapixel visualization of the documented documented artworks artworks use use the the QR QR codes codes oror followfollow thethe linkslinks below.below. ( a(a) )https://gpix.webs.upv.es/gpix/305.html https://gpix.webs.upv.es/gpix/305.html (accessed (accessed on on 19 19 August August 2021) 2021);; ((bb)) https://gpix.webs.upv.es/gpix/279.html https://gpix.webs.upv.es/gpix/279.html (accessed (accessed on on19 August 19 August 2021) 2021);; (c) (c) https://gpix.webs. upv.es/gpix/2197.htmlhttps://gpix.webs.upv.es/gpix/2197.html (accessed on 19 August (accessed 2021). on 19 August 2021).

TheThe rest of the the museum museum imaging imaging process process can can be be solved solved using using a proper a proper web web page page de- designsign and and this this may may depend depend on on the the needs needs of ofthe the researchers. researchers. In Inthis this case, case, Elementor Elementor [25 [25], a], asimple simple and and free free w webeb design design plugin plugin based based on on WordPress WordPress [26] [26],, was was used used to to develop the entireentire webpagewebpage thatthat allow allow the the results results to to be be disseminated: disseminated www.gpix.upv.es: www.gpix.upv.es (accessed (accessed on on 19 August19 August 2021). 2021). Author Contributions: Photographic capture and Image Editing, P.M.C.-B., P.R.-N. and T.G.-P.; Author Contributions: Photographic capture and Image Editing, P.M.C.-B., P.R.-N. and T.G.-P.; Topographic measurements and SfM restitutions P.R.-N. and T.G.-P.; Visualization techniques and Topographic measurements and SfM restitutions P.R.-N. and T.G.-P.; Visualization techniques and dissemination, P.M.C.-B. All authors have read and agreed to the published version of the manu- dissemination, P.M.C.-B. All authors have read and agreed to the published version of the manuscript. script. Funding: This research was carried out within the Research Project entitled Captura fotográfica de Funding: This research was carried out within the Research Project entitled Captura fotográfica de resolución gigapíxel para la documentación y divulgación del patrimonio pictórico (01/01/19–01/01/21), resolución gigapíxel para la documentación y divulgación del patrimonio pictórico (01/01/19–01/01/21), ref- Primeros Proyectos de Investigaci n (PAID-06-18), referenceerence SP20180066. SP20180066. Project Project funded funded with with the the help help of of Primeros Proyectos de Investigaciónó (PAID-06-18), VicerrectoradoVicerrectorado dede InvestigaciInvestigación,Innovaciónón, Innovación y Transferencia de la Univers Universitatitat Politècnica Politècnica de de València València (UPV), (UPV), ValValènciaència,, Spain.Spain. Acknowledgments:Acknowledgments: WeWe warmlywarmly appreciateappreciate thethe collaborationcollaboration of the Museo de Bellas Artes de Valencia (Spain)(Spain) forfor theirtheir supportsupport andand accessaccess to the artworks. We We want want to to specially specially thank thank its its current current director, director, PabloPablo GonzGonzález,ález, asas wellwell as his predecessors, Carlos Reyero and José José Ignacio Casar. Casar. We wish also to recognizerecognize andand thankthank PilarPilar Gramage,Gramage, RamRamónón MartMartínezínez and Juan Toledo forfor theirtheir valuablevaluable help.help. ConflictsConflicts ofof Interest:Interest: TheThe authorsauthors declaredeclare nono conflictconflict ofof interest.interest.

ReferencesReferences 1.1. Proctor,Proctor, N. N. The The Google Google Art Art Project: Project: A A newnew generationgeneration ofof museumsmuseums onon thethe web?web? Mus.Mus. J.J. 20112011,, 5454,, 215–221.215–221. [CrossRef] 2.2. GoogleGoogle Arts Arts & & Culture. Culture. AvailableAvailable online: online: https://artsandculture.google.com https://artsandculture.google.com (accessed (accessed on on 12 12 July July 2021). 2021). 3.3. C2RMF.C2RMF. Available Available online: online: https://c2rmf.fr https://c2rmf.fr ((accessedaccessed on on 12 12 July July 2021 2021).). 4.4. Haltadeﬁnizione.Haltadefinizione. AvailableAvailable online: online: https://www.haltadeﬁnizione.com https://www.haltadefinizione.com (accessed(accessed on on 12 12 July July 2021). 2021). 5.5. Madpixel.Madpixel. AvailableAvailable online: online: https://www.madpixel.es https://www.madpixel.es (accessed(accessed on on 12 12 July July 2021). 2021). 6.6. Goldstein,Goldstein, D.B. D.B. Physical Physical Limits limits in in Digital digital Photography. photography.Northlight Northlight Images Images. 2009. 2009 Available. Available online: online: http://www.northlight-images. http://www.northlight-im- co.uk/article_pages/guest/physical_limits.htmlages.co.uk/article_pages/guest/physical_limits.html (accessed (accessed on 19on August19 August 2021). 2021). 7.7. Quesada,Quesada, E.;E.; Harman,Harman, J.J. AA stepstep furtherfurther inin rockrock artart digitaldigital enhancements.enhancements. DStretchDStretch onon GigapixelGigapixel imaging. Digit. Appl. Archaeol. Archaeol. Cult.Cult. Herit. Herit.2019 2019,,13 13.[, doi:CrossRef10.1016/j.daach.2019.e00098] .

J. Imaging 2021, 7, 156 20 of 20

8. Cosentino, A. A practical guide to panoramic multispectral imaging. eConserv. Mag. 2013, 25, 64–73. 9. Baglioni, L.; Romor, J.; Salvatore, M. Immagini High Resolution per il rilievo delle architetture illusorie High Res Images for illusory architectures survey. Patrim. Siti UNESCO Mem. Misura Armon. 2013, 67–78. 10. Cabezos-Bernal, P.M.; Albert-Ballester, J.; Molina-Siles, P.; Martín-Fuentes, D. Filippo Fontana’s quadratura painting in the Church of Santa Maria del Temple of Valencia. Prospett. Archit. Conserv. Digit. Divulg. E Stud. 2016, 2, 65–78. 11. The Gigapixel Project. Available online: https://web.archive.org/web/20060719063806/ (accessed on 12 July 2021). 12. Kopf, J.; Uyttendaele, M.; Deussen, O.; Cohen, M. Capturing and Viewing Gigapixel Images. ACM Trans. Graph 2007, 26, 10. [CrossRef] 13. The Verge. Available online: https://www.theverge.com/2016/5/17/11686296/art-camera-google-cultural-institute (accessed on 12 July 2021). 14. Clauss. Available online: https://dr-clauss.de/en (accessed on 12 July 2021). 15. Madpixel ROB. Available online: https://www.madpixelrob.com (accessed on 12 July 2021). 16. Conrad, J. Depth of Field in Depth. Large Form. Page. 2004. Available online: https://www.largeformatphotography.info/articles/ DoFinDepth.pdf (accessed on 19 August 2021). 17. Panorama Tools. Available online: https://www.panotools.org/dersch (accessed on 12 July 2021). 18. Hugin-Panorama Photo Stitcher. Available online: http://hugin.sourceforge.net (accessed on 12 July 2021). 19. PTGui. Available online: https://www.ptgui.com (accessed on 12 July 2021). 20. RAWTherapee. Available online: https://www.rawtherapee.com (accessed on 12 July 2021). 21. Warda, J.; Frey, F.; Heller, D.; Kushel, D.; Vitale, T.; Weaver, G. The AIC Guide to Digital Photography and Conservation Documentation, 2nd ed.; American Institute for Conservation of Historic and Artistic Works: Washington, DC, USA, 2011. 22. Elkhuizen, W.S.; Dore-Callewaert, T.W.J.; Leonhardt, E.; Vandivere, A.; Song, Y.; Pont, S.C.; Geraedts, J.M.P.; Dik, J. Comparison of three 3D scanning techniques for paintings, as applied to Vermeer’s ‘Girl with a Pearl Earring’. Herit. Sci. 2019, 7, 1–22. [CrossRef] 23. Zoomify Free. Available online: http://www.zoomify.com/free.htm (accessed on 12 July 2021). 24. OpenSeadragon. Available online: https://openseadragon.github.io (accessed on 12 July 2021). 25. Elementor. Available online: https://elementor.com (accessed on 8 August 2021). 26. WordPress. Available online: https://wordpress.com (accessed on 8 August 2021).