Vrije Universiteit Brussel

An exploratory study towards objective quality evaluation of digital hologram codingtoolsCorda, Roberto; Gilles, Antonin; Oh, Kwan-Jung; Pinheiro, Antonio; Schelkens, Peter; Perra,CristianPublished in:SPIE Optics + Photonics 2019, Applications of Digital Image Processing XLII

DOI:10.1117/12.2528402

Publication date:2019

Document Version:Final published version

Link to publication

Citation for published version (APA):Corda, R., Gilles, A., Oh, K-J., Pinheiro, A., Schelkens, P., & Perra, C. (2019). An exploratory study towardsobjective quality evaluation of digital hologram coding tools. In SPIE Optics + Photonics 2019, Applications ofDigital Image Processing XLII (Vol. 11137). [111371D] SPIE. https://doi.org/10.1117/12.2528402

General rightsCopyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright ownersand it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

• Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portal

Take down policyIf you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediatelyand investigate your claim.

Download date: 23. Sep. 2020

PROCEEDINGS OF SPIE

SPIEDigitalLibrary.org/conference-proceedings-of-spie

An exploratory study towardsobjective quality evaluation of digitalhologram coding tools

Roberto Corda, Antonin Gilles, Kwan-Jung Oh, AntonioPinheiro, Peter Schelkens, et al.

Roberto Corda, Antonin Gilles, Kwan-Jung Oh, Antonio Pinheiro, PeterSchelkens, Cristian Perra, "An exploratory study towards objective qualityevaluation of digital hologram coding tools," Proc. SPIE 11137, Applications ofDigital Image Processing XLII, 111371D (6 September 2019); doi:10.1117/12.2528402

Event: SPIE Optical Engineering + Applications, 2019, San Diego, California,United States

Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 12 Sep 2019 Terms of Use: https://www.spiedigitallibrary.org/terms-of-use

An exploratory study towards objective quality evaluationof digital hologram coding tools

Roberto Cordaa, Antonin Gillesb, Kwan-Jung Ohc, Antonio Pinheirod, Peter Schelkense,f, andCristian Perraa

aDIEE, UdR CNIT, University of Cagliari, ItalybIRT b<>com, Cesson-Sevigne, France

cElectronics and Telecommunications Research Institute (ETRI), KoreadUniversidade da Beira Interior and Instituto de Telecomunicacoes, Portugal

eVrije Universiteit Brussels, Belgiumfimec, Leuven, Belgium

ABSTRACT

Holography is an acquisition and reproduction technique of visual content which allows, theoretically, for thereconstruction of the acquired scene without any difference with its real-world counterpart. The objectivequality assessment of digital holograms coding tools is a very challenging problem because the signal propertiesof holograms are significantly different from those of regular images. Several approaches can be devised forholography compression and objective quality evaluation. The exploratory study presented in this paper aimsat assessing a procedure for objective quality evaluation of data compression tools when applied to the hologramplane.

Keywords: Digital holography, JPEG Pleno, compression, quality

1. INTRODUCTION

Holography is an acquisition and reproduction technique of visual contents which allows, theoretically, for thereconstruction of the acquired scene without any difference with its real-world counterpart.1 The most popular3D visualization technologies, which often are based on the stereoscopic principle,2–4 are affected by the vergence-accommodation conflict (VAC) as major problem and difference with respect to the real world vision. On thecontrary, the VAC is absent in holography.

Recently, digital holography received a lot of attention because it simplifies acquisition, reproduction andtransmission processes when compared to analog holography. Microscopy, tomography and interferometry areexamples of the main use cases.5 In microscopy applications, digital holography can bring several benefits, asit aids to overcome the limited field of depth at high magnifications, without the need for laborious mechanicalrefocusing .6,7 In screening applications it can be used as a fast labelling-free technique, allowing for cell analysiswithout inducing alterations.8 In tomography, morphology and chemical information about the cell can beobtained studying its refractive index (RI) and high resolution 3D RI tomograms can be numerically reconstructedto obtain useful information regarding the pathophysiology of diseases.9,10 Interferometric acquisition allowsfor non-invasive measurements of displacements and/or deformations of materials. This acquisitions allow fornanometer precision depth measurements.11 This investigation technique can be applied in a wide range of fields,

Further author information: (Send correspondence to C:Perra)R. Corda: E-mail: r.corda@icloud.comA. Gilles: E-mail: antonin.gilles@b-com.comK.-J. Oh: E-mail: kjoh@etri.re.krA. Pinheiro: E-mail: pinheiro@ubi.ptP. Schelkens: peter.schelkens@vub.beC. Perra: E-mail: cperra@ieee.org

Applications of Digital Image Processing XLII, edited by Andrew G. Tescher, Touradj Ebrahimi, Proc. of SPIEVol. 11137, 111371D · © 2019 SPIE · CCC code: 0277-786X/19/$21 · doi: 10.1117/12.2528402

Proc. of SPIE Vol. 11137 111371D-1Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 12 Sep 2019Terms of Use: https://www.spiedigitallibrary.org/terms-of-use

ranging from cortical bone compression and load tests,12 to wood deformation caused by humidity13 but also toperform real-time measurements of oscillation models of large structures as buildings.14

Display applications is another context in which holography could introduce thrilling innovations. Differenttypes of holographic displays have been proposed so far,15 including head-mounted prototypes,16 which are lesspenalized from a technical point of view, due to the limited display size and the short distance between the displayand the observers eyes. Head-mounted devices could therefore be the first commercially available holographicdisplays in the near future, while bigger high quality multi-user displays are not supposed to be available in thenear future.17

Different research challenges still need to be addressed, and data compression is one of those: digital hologramsare characterized by a vast amount of data, having several different features compared to classical image andvideo content. Data compression is thus an aspect of strategic importance in the development and deploymentof holography as imaging technique. Standard lossless compression algorithms have been employed,18,19 anddifferent quantization methods have been compared20 on optically acquired holograms. Transforms that aresuccessfully employed on standard image and video contents, such as the Discrete Wavelet Transform21 and theDiscrete Cosine Transform,22 have been applied also on digital holograms, as well as other types of transformssuch as Fresnelets23 and Gabor wavelets,24,25 better suited for holographic contents, have been proposed andtested on digital holographic data. Standard still image and video codecs have been also evaluated26,27 as well asmodifications to the standard codec pipelines.26,28 To improve rate-distortion performance, the hologram codingafter the back-propagation has been investigated29,30 and relying on this concept, recently a new lossless integerFresnel transform method has been proposed and tested on computer generated holograms.31 Moreover, machinelearning based methods have been proposed: a convolutional neural network has been developed to enhanceholograms after the compression with the JPEG codec,32 but also an entire deep neural network compressionand decompression scheme has been proposed.31

A procedure for objective quality evaluation of digital hologram coding tools is presented and discussed. Theprocedure considers a basic setting where the hologram plane is the input to the coding process. Signal to noiseration (SNR) on the hologram plane and average peak SNR (PSNR) on the recontructed images are chosenas objective quality metrics. Some digital holographic samples are selected from the JPEG Pleno database asreference data for testing the overall procedure. The experimental evaluation takes into consideration severalcoding tools such as JPEG 2000, H.265/HEVC, Google’s VP9 and AOMedia Video 1 (AV1).

The paper is organized as follows. The JPEG Pleno dataset is discussed in Section 2. The proposed frameworkfor objective quality evalution of digital holography compression is presented in Section 3. Section 4 reports theexperimental analysis. Finally, Section 5 draws the conclusions.

2. JPEG PLENO DATASET

To evaluate the performance of coding technologies it is of utmost importance to define a test data set that issufficiently diverse and challenging such that the targeted application domains are sufficiently covered in termsof content behavior and consequently the investigated coding technologies effectively stress-tested. Hence, inthe context of the JPEG Pleno standardization 33 effort of which it is the intention to produce among others a

Table 1: JPEG Pleno datasets summaryDataset Acquisition Target application Data format Data type

B-Com Repository CGH Display Complex field8 bit unsigned /

32 bit floatEmergImg-HoloGrail

(v1, v2)Optical Display

Complex field /Interferogram (v1)

64 bit float

ERC Interfere(I, II, III)

CGH Display Complex field 32 bit float

Tomocube Optical Tomography Complex field 32 bit floatWUT Display Optical Display Interferogram 8 bit unsigned

WUT Microscopy Optical Microscopy Interferogram 8 bit unsigned

Proc. of SPIE Vol. 11137 111371D-2Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 12 Sep 2019Terms of Use: https://www.spiedigitallibrary.org/terms-of-use

specification for holographic content coding, the JPEG committee created the JPEG Pleno Database,34 whichis currently composed of the following eight datasets: B-Com Repository,35,36 ERC Interfere I,37 II38 and III,39

Tomocube,40,41 EmergImg-HoloGrail v1 and v2,30 WUT Display42 and WUT Microscopy.43,44 The hologramshave been generated with different methods and they target different application fields. In particular, B-Com,EmergImg-HoloGrail, ERC Interfere, and WUT Display are dedicated to display applications, while Tomocubeand WUT Microscopy target microscopy and tomography applications. The database comprises 63 still samplesand 7 video sequences (the latter belonging to B-Com Repository and Emerg-Img Holograil v1). The B-ComRepository and ERC Interfere samples are Computer Generated Holograms (CGH), for a total of 45 CGH. TheEmergImg-HoloGrail, Tomocube and WUT Display/Microscopy have been optically acquired, for a total of 18optically acquired samples.

The holographic database created by the Institute of Research and Technology B-Com is composed of differentCGH, generated with two different algorithms. The dataset comprises 21 holograms acquired from syntheticscenes and have resolutions from 1920 × 1080 to 16384 × 16384 with pixel pitch from 6.4µm to 0.4µm. Twovideo sequences at different resolutions of real scenes are also proposed.

Interfere I comprises both monochromatic holograms generated from 2D and 3D objects and produced withthe multiple wavefront recording plane CGH method with occlusion culling.45 All the holograms have a pixelpitch equal to 8µm and a resolution equal to 1920 × 1080, except for the sample 3D Cat, whose pixel pitch isequal to 2µm and it has resolution of 8192 × 8192. Improving the algorithm of Interfere I, the authors haveproposed Interfere II, adding the simulation of diffuse light reflection on surfaces. The data set comprises a totalof 12 monochromatic samples. The resolution is 8192 × 8192 with a pixel pitch of 1µm for all holograms. Themost recent Interfere III data base, consists of 6 color CGH. The generation algorithm based on45 includes anextension of the Phong illumination model that allows ambient, diffuse and specular reflection support.46 Theholograms have resolutions between 1920 × 1080 and 16384 × 16384, with pixel pitches ranging from 8µm to1µm.

Other than computer-generated holograms, the JPEG Pleno Database includes also some examples of opti-cally acquired datasets of macroscopic objects. The EmergImg-HoloGrail is one of of them, and each hologram isderived from the acquisition and combination of four interferograms in order to retrieve phase information. It isorganized in two sets of monochromatic holograms. The first set (v1) includes three holograms with a resolutionof 972 × 972 and 4.4µm of pixel pitch, and a video sequence consisting of 54 holograms at 600 × 600 resolution.The second set (v2) comprises six 2588 × 2588 samples with pixel pitch equal to 2.2µm.

The WUT Display dataset includes one monochromatic and one color hologram, acquired with a syntheticaperture technique. The former has resolution of 19794 × 2020 while the latter’s resolution is 5394 × 2016, andboth have a pixel pitch of 3.45µm.

As it concerns microscopy applications, the JPEG Pleno Database comprises the WUT Microscopy dataset,recorded with an off-axis configuration and composed of three holograms with resolution 2464 × 2056 and pixelpitch of 3.45µm. The last dataset currently included in the JPEG Pleno Database is released by Tomocube, andcomprises four monochromatic samples with resolution from 512 × 512 to 24478 × 16641 and pixel pitch from0.08µm to 0.366µm.

In the B-Com Repository, ERC Interfere, EmergImg-HoloGrail and Tomocube datasets, the samples areprovided as complex matrices. The WUT holograms are represented as interferograms, and also EmergImg-HoloGrail v1 provides the interferograms from which the complex representation is derived. The datatype ofWUT samples is 8 bit unsigned, while the Interfere samples are expressed in 32 bit unsigned float. B-comRepository samples are provided both in 8 bit unsigned and in 32 bit float. Finally, the complex representationof EmergImg-HoloGrail samples is expressed in 64 bit float type. The data set features described above aresummarized in Table 1, while in Fig. 1 some examples of reconstructed holograms of these databases are provided.

Proc. of SPIE Vol. 11137 111371D-3Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 12 Sep 2019Terms of Use: https://www.spiedigitallibrary.org/terms-of-use

(a) (b) (c) (d) (e)

Figure 1: B-Com Repository: Specular Car 16k (a), EmergImg-Holograil v2: Astronaut (b), Interfere III: Biplane16k (c), WUT Microscopy: Selected HeLa Cells (d) and WUT Display: Lowiczanka Doll (e).

3. DIGITAL HOLOGRAPHY COMPRESSION

In this Section, the framework employed for the objective quality evaluation and comparison of different codingtools on digital holograms is described. The codecs under test are AOMedia Video 1 (AV1) 0.1.0,47 H.265/HEVC(reference software 16.18),48 VP9 (FFMpeg 4.0)49 and JPEG 2000 (Kakadu Software 7.10.2).50 The first threeare video codecs, while the last is a still image codec. The scheme of the framework is reported in Fig. 2 andwill be described in the hereafter.

Rangemapping &

Quantization

Re (8 bps)

Im (8 bps)

Re coded

Im coded

Re dec. (8 bps)Re (8 bps)

Im (8 bps) Im dec. (8 bps)

Im dec. (8 bps)

Re dec. (8 bps)

Reconstructed image (reference) Reconstructed image (test)

Complex data(raw)

Rendering Rendering

codec parameters8 bps

rendering parameters rendering parameters

Hologramplane data

SNR

Re

Im

PSNR

Components to complex

Complex data(decoded)

Complex data(raw)

Complex to components

Coding Decoding

Figure 2: Compression architecture.

The holograms that have been employed in the compression experiments are solely CGHs that are repre-sented by a matrix, in which each element is a complex number. Standard codecs typically do not supportthis type of representation as input data, thus pre and post processing operations are performed on the databefore the compression stage and after the decompression stage. The Hologram plane data block is the two-dimensional array of complex values, from which the real (Re) and imaginary (Im) components are calculatedwithin the Complex to components block. Although the complex data could be converted in other forms (suchas the amplitude-phase format), the real-imaginary format has been chosen since it is known to provide bettercompression performances.17,25,26,30 The real and imaginary components, are both mapped in the range [0, 1]in the Range mapping & Quantization block, and then each component is quantized to 8 bit per sample (bps),resulting in the range [0, 255]. This step was included to provide a compatible input to all codecs under test.In the Coding block, the two components are independently encoded with each codec at seven different qualitylevels. Subsequently, the Decoding block performs the decoding of the coded components. At this stage, the

Proc. of SPIE Vol. 11137 111371D-4Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 12 Sep 2019Terms of Use: https://www.spiedigitallibrary.org/terms-of-use

effects of the compression are evaluated through the SNR metric, defined as:51

SNR = 10 log10

M−1∑x=0

N−1∑y=0

[r(x, y)]2

M−1∑x=0

N−1∑y=0

[r(x, y) − t(x, y)]2(1)

where r(x, y) is the reference data, while t(x, y) denotes the data under test, in order to compare the twocomponents before and after the coding process.

Subsequently, the complex matrix that represents the hologram is restored through the Components to com-plex block, and the holographic image is reconstructed in the Rendering block in a predetermined reconstructionplane, aka object plane. This operation is performed on the hologram that has been compressed, representedby Complex data (decoded), but also on the original hologram represented by Complex data (raw). The tworeconstructed images are finally compared through the PSNR metric:

PSNR = 10 log10

(2n − 1)2

1

MN

M−1∑x=0

N−1∑y=0

[r(x, y) − t(x, y)]2(2)

For each quality level of the codecs, the PSNR evaluation is performed (when allowed by the hologram undertest) on several reconstructions, carried out at different depths and different viewpoints.52

4. EXPERIMENTAL ANALYSIS

In this Section, the experimental results obtained applying the compression scheme described in Section 3 areshown. In this paper, we limit the reporting to a subset of holograms selected from the aforementioned datasets.The holograms under test are 3D Multi from Interfere I, Ball 8KD from Interfere II, Venus from Interfere IIIand Specular Car 8K from B-Com Repository.

For each hologram, the compression results of the real and imaginary components (in terms of SNR) and theobjective quality assessment in the reconstructed planes (in terms of PSNR) are reported. The four hologramsunder test allow for the image reconstruction at different distances, which fall in a specific interval defined by thehologram itself. Different portions of the scene can be focused by varying the reconstruction distance. Taking thisfeature into account, for Ball 8KD, Venus and Specular Car 8K, the PSNR is thus evaluated, for every qualitylevel of the codecs, at 30 different reconstruction distances. These reconstruction distances cover uniformlythe specific reconstruction distance range allowed by each hologram. The graphs below show the mean PSNRobtained with these 30 different reconstructions, along with the standard deviation. For the 3D Multi holograminstead, the authors recommend three different reconstruction distances, each of which brings in focus one ofthe three dices depicted in the holographic image, therefore for this hologram only three different reconstructiondistances are evaluated and averaged.

The Ball 8KD and Specular Car 8K holograms allow also for the reconstruction at different viewpoints,therefore for these two holograms, the PSNR is evaluated at 36 different reconstruction viewpoints, for everyquality level of the codecs. The 36 different viewpoints have been chosen in order to uniformly cover the range ofdifferent viewpoints allowed by the specific hologram under test. Also in this case, the mean PSNR along withits standard deviation is reported in the graphs.

In Figs. 3–4 the compression results obtained respectively with 3D Multi and Ball 8KD are shown. Bothsamples are gray scale; however, the former does not support reconstruction from different viewpoints, as opposedto the second. In terms of components, it can be noted as the real (Fig. 3a and Fig. 4a), and the imaginary part(Fig. 3b. Fig. 4b) shows almost identical results on both holograms, so they react in a very similar manner tocompression. JPEG 2000 is performing the poorest on both samples, for bitrates up to 2.5bps on 3D Multi andup to 4bps on Ball 8KD. For bitrate values close to zero and up to 1.9bps, in both holograms the performancedifference among the video codecs is negligible, while above 1.9bps it can be noted the superior performances of

Proc. of SPIE Vol. 11137 111371D-5Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 12 Sep 2019Terms of Use: https://www.spiedigitallibrary.org/terms-of-use

HEVC and AV1, on both samples, while the distortion of VP9 saturates as the bitrate increases. On the contrary,JPEG 2000 increases gradually its performance as the bitrate increases. With regard to the average quality of thereconstructed image with the variation of reconstruction distance, the results on 3D Multi are reported in Fig. 3c.Each point in the graph represents the mean PSNR value, while the vertical interval represents the standarddeviation. As could be expected from the component analysis, the video codecs are the best up to 4.5bps, whilethe JPEG 2000 is the worst. Beyond this value, the difference between AV1, HEVC and JPEG 2000 is negligible,while VP9 is the worst. The average quality of the reconstructed image with the variation of reconstructiondistance for Ball 8KD is reported in Fig. 4c. It can be noted that also in this case the results are consistent withthe results obtained in terms of components, with the video codecs that provide the best performances at lowbitrates, while JPEG 2000 improves at higher bitrates.

Similar observations can be made analyzing the image quality with the change in viewpoint (Fig. 4d). ThePSNR standard deviation is slightly higher at medium-low and at high bitrates. However the highest standarddeviation values never exceed ±2dB. It can also be noted that Ball 8KD, at the same bitrates, shows lower SNRand PSNR values than 3D Multi. Ball 8KD is a ”more advanced” sample, that supports diffuse reflections andfull parallax effects, and these features affect also the compression performances.

Fig. 5 and Fig. 6 show the results obtained respectively from the Venus and Specular Car 8K samples. Alsoin this case, only the second hologram supports reconstructions from different viewpoints; however, in this casethe two samples are RGB. It can be noted that even on these two samples, the real and imaginary parts react inthe same way on the compression process. On Venus (Fig. 5a–5b) the performances of the four codecs are similarup to 11bps, while above this threshold HEVC and JPEG 2000 have the best performances. The worst codecis VP9, while the most recent AV1 has lower performances at high bitrates, if compared to HEVC and JPEG2000. The results of Specular Car 8K in terms of components (Fig. 6a–6b) show instead than up to 4bps thebest codecs are the most recent (and complex) AV1 and HEVC. However, beyon this threshold, AV1 does notmaintain performances similar to HEVC, but on the contrary its performances become similar to the worst codec,VP9, as the bitrate increases. JPEG 2000 also in this case improves its performances as the bitrate increases,but surpasses HEVC only at high bitrates, close to 14bps. In both samples the overall PSNR trends (Fig. 5c andFig. 6c–6d) are consistent with those of the SNR in terms of components, thus the same consideration apply.For what concerns the results obtained with Venus in terms of mean PSNR with the change in reconstructiondistance, it can be noted that the standard deviation is near zero for all codecs: there are not appreciablevariations in image quality with varying distances. However, the Specular Car 8K shows a different behavior.Both with the variation in the reconstruction distance (Fig. 6c) and in the viewpoint (Fig. 6d) the standarddeviation is not equal to zero and it shows a gradual reduction with the increase in the bitrate. Also in the worstcases, which occur especially at medium-low bitrates, it never exceeds ±2dB, similar to 3D Multi and Ball 8KD.Variations of that entity should be rarely perceptible by the observer.

In some cases it can be noted an unpredictable trend of the Rate-Distortion curves, as in Ball 8KD, with thevariation in the reconstruction distance with the JPEG 2000 codec (Fig. 4c). This fact can be caused by the

0 2 4Rate [bps]

10

20

30

40

50

60

SN

R [

dB

]

HEVCJPEG2000VP9AV1

(a)

0 2 4Rate [bps]

10

20

30

40

50

60

SN

R [

dB

]

HEVCJPEG2000VP9AV1

(b)

0 2 4 6 8Rate [bps]

10

20

30

40

50

60

PS

NR

[d

B]

HEVCJPEG2000VP9AV1

(c)

Figure 3: 3D Multi results: SNR real part (a), SNR imaginary part (b), mean PSNR at different reconstructiondistances (c).

Proc. of SPIE Vol. 11137 111371D-6Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 12 Sep 2019Terms of Use: https://www.spiedigitallibrary.org/terms-of-use

0 2 4 6Rate [bps]

10

20

30

40

50

60

SN

R [

dB

]

HEVCJPEG2000VP9AV1

(a)

0 2 4 6Rate [bps]

10

20

30

40

50

60

SN

R [

dB

]

HEVCJPEG2000VP9AV1

(b)

0 5 10 15Rate [bps]

10

20

30

40

50

60

PS

NR

[d

B]

HEVCJPEG2000VP9AV1

(c)

0 5 10 15Rate [bps]

10

20

30

40

50

60

PS

NR

[d

B]

HEVCJPEG2000VP9AV1

(d)

Figure 4: Ball 8KD results: SNR real part (a), SNR imaginary part (b), mean PSNR at different reconstructiondistances (c), mean PSNR from different viewpoints (d).

lack of data content-awareness of the codecs: the modifications of the standard lossy coding (that cannot takesthe hologram features into account) can have unpredictable effects of the reconstructed image quality.

Finally, it should be noted that the process of hologram rendering into the object plane originates the so calledspeckle noise which drastically affects the PSNR measures. The influence of speckle noise on the reconstructedimage quality is an open research problem.

0 10 20Rate [bps]

10

20

30

40

50

60

SN

R [

dB

]

HEVCJPEG2000VP9AV1

(a)

0 10 20Rate [bps]

10

20

30

40

50

60

SN

R [

dB

]

HEVCJPEG2000VP9AV1

(b)

0 20 40Rate [bps]

0

10

20

30

40

50

60

PS

NR

[d

B]

HEVCJPEG2000VP9AV1

(c)

Figure 5: Venus results: SNR real part (a), SNR imaginary part (b), mean PSNR at different reconstructiondistances (c).

0 5 10 15Rate [bps]

10

20

30

40

50

60

SN

R [

dB

]

HEVCJPEG2000VP9AV1

(a)

0 5 10 15Rate [bps]

10

20

30

40

50

60

SN

R [

dB

]

HEVCJPEG2000VP9AV1

(b)

0 10 20 30Rate [bps]

10

20

30

40

50

60

PS

NR

[d

B]

HEVCJPEG2000VP9AV1

(c)

0 10 20 30Rate [bps]

10

20

30

40

50

60

PS

NR

[d

B]

HEVCJPEG2000VP9AV1

(d)

Figure 6: Specular Car 8K results: SNR real part (a), SNR imaginary part (b), mean PSNR at differentreconstruction distances (c), mean PSNR from different viewpoints (d).

Proc. of SPIE Vol. 11137 111371D-7Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 12 Sep 2019Terms of Use: https://www.spiedigitallibrary.org/terms-of-use

5. CONCLUSIONS

A digital hologram is a representation of the plenoptic function, which carries more information of a conventionaldigital photo. It allows for the reconstruction of a scene at different viewing angles and reconstruction planes. Adigital hologram can correspond to a considerable amount of samples which creates a demand for compressionwhen being used in practical applications. This paper presents an objective quality evaluation framework. Itshould be noted that the compression is performed naively in the hologram plane and that the codecs under testhave not been tuned to operate in this domain. The sole purpose of this experiment is to evaluate the overallprocedure. More advanced settings can also include backprojection operation to allow for compression in thereconstruction plane and/or deployment of more specialized codecs that are better suited to handle holographiccontent.

ACKNOWLEDGMENTS

The research leading to these results received funding from the Cagliari2020 project (MIUR, PON04a2 00381),the DigitArch Cluster Top-Down project (POR FESR, 2014-2020), the European Research Council under theEuropean Unions Seventh Framework Programme (FP7/2007- 2013)/ERC Grant Agreement n. 617779 (IN-TERFERE), and the Giga-KOREA project (GK19D0100, Development of Telecommunications Terminal withDigital Holographic Table-top Display).

REFERENCES

[1] Slinger, C., Cameron, C., and Stanley, M., “Computer-generated holography as a generic display technol-ogy,” Computer 38(8), 46–53 (2005).

[2] Konrad, J. and Halle, M., “3-D displays and signal processing,” IEEE Signal Processing Magazine 24(6),97–111 (2007).

[3] Reichelt, S., Haussler, R., Leister, N., Futterer, G., Stolle, H., and Schwerdtner, A., “Holographic 3-Ddisplays-electro-holography within the grasp of commercialization,” in [Advances in lasers and electro optics ],IntechOpen (2010).

[4] Cuervo, E., “Beyond reality: Head-mounted displays for mobile systems researchers,” GetMobile: MobileComputing and Communications 21(2), 9–15 (2017).

[5] Schelkens, P., Ebrahimi, T., Gilles, A., Gioia, P., Oh, K.-J., Pereira, F., Perra, C., and Pinheiro, A. M.,“Jpeg pleno: Providing representation interoperability for holographic applications and devices,” ETRIJournal 41(1), 93–108 (2019).

[6] Mann, C. J., Yu, L., and Kim, M. K., “Movies of cellular and sub-cellular motion by digital holographicmicroscopy,” BioMedical Engineering OnLine 5(1), 21 (2006).

[7] Kemper, B., Bauwens, A., Vollmer, A., Ketelhut, S., Langehanenberg, P., Muthing, J., Karch, H., andvon Bally, G., “Label-free quantitative cell division monitoring of endothelial cells by digital holographicmicroscopy,” Journal of Biomedical Optics 15(3), 036009 (2010).

[8] Kuhn, J., Shaffer, E., Mena, J., Breton, B., Parent, J., Rappaz, B., Chambon, M., Emery, Y., Magistretti,P., Depeursinge, C., et al., “Label-free cytotoxicity screening assay by digital holographic microscopy,”Assay and drug development technologies 11(2), 101–107 (2013).

[9] Kim, K., Yoon, H., Diez-Silva, M., Dao, M., Dasari, R. R., and Park, Y., “High-resolution three-dimensionalimaging of red blood cells parasitized by plasmodium falciparum and in situ hemozoin crystals using opticaldiffraction tomography,” Journal of biomedical optics 19(1), 011005 (2013).

[10] Koo, S.-e., Jang, S., Park, Y., and Park, C. J., “Reconstructed three-dimensional images and param-eters of individual erythrocytes using optical diffraction tomography microscopy,” Annals of laboratorymedicine 39(2), 223–226 (2019).

[11] Kim, M. K., “Principles and techniques of digital holographic microscopy,” SPIE reviews 1(1), 018005(2010).

[12] Ruiz, C. G. T., Manuel, H., Flores-Moreno, J., Frausto-Reyes, C., and Santoyo, F. M., “Cortical bonequality affectations and their strength impact analysis using holographic interferometry,” Biomedical OpticsExpress 9(10), 4818–4833 (2018).

Proc. of SPIE Vol. 11137 111371D-8Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 12 Sep 2019Terms of Use: https://www.spiedigitallibrary.org/terms-of-use

[13] Kumar, M. and Shakher, C., “Experimental characterization of the hygroscopic properties of wood duringconvective drying using digital holographic interferometry,” Applied optics 55(5), 960–968 (2016).

[14] Poggi, P., Locatelli, M., Pugliese, E., Delle Donne, D., Lacanna, G., Meucci, R., and Ripepe, M., “Remotemonitoring of building oscillation modes by means of real-time mid infrared digital holography,” Scientificreports 6, 23688 (2016).

[15] Pan, Y., Liu, J., Li, X., and Wang, Y., “A review of dynamic holographic three-dimensional display:algorithms, devices, and systems,” IEEE Transactions on Industrial Informatics 12(4), 1599–1610 (2015).

[16] Maimone, A., Georgiou, A., and Kollin, J. S., “Holographic near-eye displays for virtual and augmentedreality,” ACM Transactions on Graphics (TOG) 36(4), 85 (2017).

[17] Blinder, D., Ahar, A., Bettens, S., Birnbaum, T., Symeonidou, A., Ottevaere, H., Schretter, C., andSchelkens, P., “Signal processing challenges for digital holographic video display systems,” Signal Processing:Image Communication 70, 114–130 (2019).

[18] Naughton, T. J., Frauel, Y., Javidi, B., and Tajahuerce, E., “Compression of digital holograms for three-dimensional object reconstruction and recognition,” Applied Optics 41(20), 4124–4132 (2002).

[19] Shortt, A. E., Naughton, T. J., and Javidi, B., “A companding approach for nonuniform quantization ofdigital holograms of three-dimensional objects,” Optics Express 14(12), 5129–5134 (2006).

[20] Xing, Y., Pesquet-Popescu, B., and Dufaux, F., “Comparative study of scalar and vector quantization ondifferent phase-shifting digital holographic data representations,” in [2014 3DTV-Conference: The TrueVision-Capture, Transmission and Display of 3D Video (3DTV-CON) ], 1–4, IEEE (2014).

[21] Shortt, A. E., Naughton, T. J., and Javidi, B., “Compression of digital holograms of three-dimensionalobjects using wavelets,” Optics Express 14(7), 2625–2630 (2006).

[22] Seo, Y.-H., Choi, H.-J., and Kim, D.-W., “3d scanning-based compression technique for digital hologramvideo,” Signal Processing: Image Communication 22(2), 144–156 (2007).

[23] Liebling, M., Blu, T., and Unser, M., “Fresnelets: new multiresolution wavelet bases for digital holography,”IEEE Transactions on image processing 12(1), 29–43 (2003).

[24] Viswanathan, K., Gioia, P., and Morin, L., “Wavelet compression of digital holograms: towards a view-dependent framework,” in [Applications of Digital Image Processing XXXVI ], 8856, 88561N, InternationalSociety for Optics and Photonics (2013).

[25] El Rhammad, A., Gioia, P., Gilles, A., Cagnazzo, M., and Pesquet-Popescu, B., “Color digital hologramcompression based on matching pursuit,” Applied optics 57(17), 4930–4942 (2018).

[26] Peixeiro, J. P., Brites, C., Ascenso, J., and Pereira, F., “Holographic data coding: Benchmarking andextending hevc with adapted transforms,” IEEE Transactions on Multimedia 20(2), 282–297 (2017).

[27] Corda, R., Giusto, D., Liotta, A., Song, W., and Perra, C., “Recent advances in the processing and renderingalgorithms for computer-generated holography,” Electronics 8(5), 556 (2019).

[28] Blinder, D., Bruylants, T., Ottevaere, H., Munteanu, A., and Schelkens, P., “Jpeg 2000-based compressionof fringe patterns for digital holographic microscopy,” Optical Engineering 53(12), 123102 (2014).

[29] Darakis, E. and Soraghan, J. J., “Reconstruction domain compression of phase-shifting digital holograms,”Applied Optics 46(3), 351–356 (2007).

[30] Bernardo, M. V., Fernandes, P., Arrifano, A., Antonini, M., Fonseca, E., Fiadeiro, P. T., Pinheiro, A. M.,and Pereira, M., “Holographic representation: Hologram plane vs. object plane,” Signal Processing: ImageCommunication 68, 193–206 (2018).

[31] Blinder, D. and Schelkens, P., “Integer fresnel transform for lossless hologram compression,” in [2019 DataCompression Conference (DCC) ], 389–397, IEEE (2019).

[32] Jiao, S., Jin, Z., Chang, C., Zhou, C., Zou, W., and Li, X., “Compression of phase-only holograms withjpeg standard and deep learning,” Applied Sciences 8(8), 1258 (2018).

[33] Ebrahimi, T., Foessel, S., Pereira, F., and Schelkens, P., “JPEG Pleno: Toward an efficient representationof visual reality,” IEEE MultiMedia 23, 14–20 (Oct 2016).

[34] “JPEG Pleno Database.” https://jpeg.org/plenodb/. (Online; accessed on 1 July 2019).

[35] Gilles, A., Gioia, P., Cozot, R., and Morin, L., “Hybrid approach for fast occlusion processing in computer-generated hologram calculation,” Applied optics 55(20), 5459–5470 (2016).

Proc. of SPIE Vol. 11137 111371D-9Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 12 Sep 2019Terms of Use: https://www.spiedigitallibrary.org/terms-of-use

[36] Gilles, A., Gioia, P., Cozot, R., and Morin, L., “Computer generated hologram from multiview-plus-depthdata considering specular reflections,” in [2016 IEEE International Conference on Multimedia & Expo Work-shops (ICMEW) ], 1–6, IEEE (2016).

[37] Blinder, D., Ahar, A., Symeonidou, A., Xing, Y., Bruylants, T., Schreites, C., Pesquet-Popescu, B., Dufaux,F., Munteanu, A., and Schelkens, P., “Open access database for experimental validations of holographic com-pression engines,” in [2015 Seventh International Workshop on Quality of Multimedia Experience (QoMEX) ],1–6, IEEE (2015).

[38] Symeonidou, A., Blinder, D., Ahar, A., Schretter, C., Munteanu, A., and Schelkens, P., “Speckle noisereduction for computer generated holograms of objects with diffuse surfaces,” in [Optics, Photonics andDigital Technologies for Imaging Applications IV ], 9896, 98960F, International Society for Optics andPhotonics (2016).

[39] Symeonidou, A., Blinder, D., and Schelkens, P., “Colour computer-generated holography for point cloudsutilizing the phong illumination model,” Optics express 26(8), 10282–10298 (2018).

[40] Shin, S., Kim, K., Yoon, J., and Park, Y., “Active illumination using a digital micromirror device forquantitative phase imaging,” Optics letters 40(22), 5407–5410 (2015).

[41] Yoon, J., Kim, K., Park, H., Choi, C., Jang, S., and Park, Y., “Label-free characterization of white bloodcells by measuring 3d refractive index maps,” Biomedical optics express 6(10), 3865–3875 (2015).

[42] Go los, A., Zaperty, W., Finke, G., Makowski, P., and Kozacki, T., “Fourier rgb synthetic aperture colorholographic capture for wide angle holographic display,” in [Optics and Photonics for Information ProcessingX ], 9970, 99701E, International Society for Optics and Photonics (2016).

[43] Stepien, P., Bernas, T., Krauze, W., Sas-Nowosielska, H., and Kujawinska, M., “Multi-modal quantitativeanalysis of hela cells using digital holographic microscopy and confocal laser scanning microscopy,” in [Speckle2018: VII International Conference on Speckle Metrology ], 10834, 108341V, International Society for Opticsand Photonics (2018).

[44] Baczewska, M., Kujawinska, M., Skrzypek, E., and Sladowski, D., “Feasibility study of investigation ofskin at cellular level by digital holographic microscopy,” in [Speckle 2018: VII International Conference onSpeckle Metrology ], 10834, 108341P, International Society for Optics and Photonics (2018).

[45] Symeonidou, A., Blinder, D., Munteanu, A., and Schelkens, P., “Computer-generated holograms by multiplewavefront recording plane method with occlusion culling,” Optics Express 23, 22149–22161 (Aug. 2015).WOS:000362418300055.

[46] Symeonidou, A., Blinder, D., and Schelkens, P., “Colour computer-generated holography for point cloudsutilizing the Phong illumination model,” Optics Express 26, 10282–10298 (Apr. 2018).

[47] “AOMedia Video 1 software.” https://aomedia.org/. (Online; accessed on 1 July 2019).

[48] “HEVC Reference Software.” https://hevc.hhi.fraunhofer.de/. (Online; accessed on 1 July 2019).

[49] “FFMpeg Software.” https://www.ffmpeg.org/. (Online; accessed on 1 July 2019).

[50] “Kakadu Software.” https://kakadusoftware.com/. (Online; accessed on 1 July 2019).

[51] Gonzalez, R. and Woods, R., [Digital Image Processing ], Pearson/Prentice Hall (2008).

[52] Oh, K. and Kim, J., “JPEG Pleno Holography 3D volumetric quality assessment.” Doc. ISO/IEC JTC1/SC 29/WG1 M78045, 78th Meeting, Rio de Janeiro, Brazil (2018).

Proc. of SPIE Vol. 11137 111371D-10Downloaded From: https://www.spiedigitallibrary.org/conference-proceedings-of-spie on 12 Sep 2019Terms of Use: https://www.spiedigitallibrary.org/terms-of-use