Tone Compatibility between HDR Displays

a footage taken during the French stage of the Volvo Ocean Race. This was done under a collaborative project by French Multimedia companies1819. 2.3 H...

0 downloads 155 Views 5MB Size
Tone Compatibility between HDR Displays Cambodge Bista , R´emi Cozota,b , G´erard Madeca , and Xavier Duclouxa a

b

IRT b<>com, 1219 Avenue des Champs Blancs, 35510 Cesson-S´evign´e, France IRISA, University of Rennes 1, 263 Avenue du General Leclerc, 35000, Rennes, France ABSTRACT

High Dynamic Range (HDR) is the latest trend in television technology and we expect an influx of HDR capable consumer TVs in the market. Initial HDR consumer displays will operate on a peak brightness of about 500-1000 nits while in the coming years display peak brightness is expected to go beyond 1000 nits. However, professionally graded HDR content can range from 1000 to 4000 nits. As with Standard Dynamic Range (SDR) content, we can expect HDR content to be available in variety of lighting styles such as low key, medium key and high key video. This raises concerns over tone-compatibility between HDR displays especially when adapting to various lighting styles. It is expected that dynamic range adaptation between HDR displays uses similar techniques as found with tone mapping and tone expansion operators. In this paper, we survey simple tone mapping methods of 4000 nits color-graded HDR content for 1000 nits HDR displays. We also investigate tone expansion strategies when HDR content graded in 1000 nits is displayed on 4000 nits HDR monitors. We conclude that the best tone reproduction technique between HDR displays strongly depends on the lighting style of the content. Keywords: High Dynamic Range, Tone Mapping, Tone Expansion

1. INTRODUCTION An obvious difference a customer can spot when walking into a consumer electronics store is that the same video content looks slightly different on various TV sets. One of the main reasons for this discrepancy is that not all consumer TV’s are configured with the same system gamma. A gamma of 2.2 fits well for ambient lighting conditions while a gamma of 2.4 gives slightly more contrast. This gamma correction has a direct effect on the tone curve and hence the same video looks slightly different on such displays. Tone compatibility is a common problem amongst many commercial Standard Dynamic Range (SDR) displays. This issue is far more critical for High Dynamic Range (HDR) television. HDR TV is the latest generation of displays with high peak brightness and an increase in contrast levels between dark and bright regions. Recently, HDR displays have attracted a lot of attention in various technology shows (CES, NAB, IBC etc.) and have also started infiltrating the consumer market. With the arrival of these new displays it is a rising concern for content providers to be able to reproduce their artistic intent on a variety of displays especially in terms of tonal compatibility. In a typical video chain, content providers professionally grade their HDR content on a mastering display ranging from 1000 to 4000 nits. This content would be distributed to consumer displays ranging from 500-1000 nits. Backward compatibility of HDR content for SDR displays has been an on going research issue. The process of HDR to SDR conversion is known as tone mapping.1 Also, forward compatibility of legacy SDR video for HDR displays has also been gaining traction in the form of tone expansion.2 Yet no study has been presented comparing reproduction of dynamic range between HDR displays. We define the nomenclature of these processes which have been defined in Table 1. In addition to this, few studies have considered the significance of complex lighting styles (extreme variation in luminance and contrast). In this paper, we set out to identify basic tone reproduction techniques between HDR displays put in context to this complex HDR video ecosystem. Further author information: (Send correspondence to Cambodge Bist) C. Bist: E-mail: [email protected] R. Cozot: E-mail: [email protected] G. Madec: E-mail: [email protected] X. Ducloux: E-mail: [email protected]

Display Content SDR Legacy Video 1000 nits Video 4000 nits Video

SDR Legacy Display

1000 nits Display

4000 nits Display

— Tone Mapping Tone Mapping

Tone Expansion — Tone Mapping

Tone Expansion Tone Expansion —

Table 1. Tone Reproduction Methods: This table explains tone compatibility terminologies between content and displays. The lines “—” represent no compatibility issues, as content has been mastered for this display. Tone Mapping is performed to reduce dynamic range of 1000 or 4000 nits content on SDR displays. Tone Expansion is used to expand the dynamic range of SDR content for HDR displays. The focus of this paper is exploring Tone Mapping and Tone Expansion methods highlighted in blue for tonal compatibility between HDR displays.

The main contribution of this paper is a perceptual evaluation of basic tone reproduction methods between HDR monitors. The evaluation consists surveying simple tone mapping methods for 4000 nits color-graded content for 1000 nits HDR displays and also investigating tone expansion strategies when content graded in 1000 nits is displayed on 4000 nits monitors. We evaluate these tone reproducing methods on content of various lighting styles such as low key, medium key and high key. Based on the results of these experiments we identify critical issues that need to be resolved for accurate tone rendering between HDR displays.

2. BACKGROUND In this section, we discuss related work on dynamic range tone reproduction. We define lighting style aesthetics to evaluate HDR artistic intent. Finally, we describe the current HDR ecosystem in terms of variety of HDR content and display.

2.1 Dynamic Range Tone Reproduction From decades of research in tone mapping we have realized that displaying a tone mapped image on an SDR display doesn’t always recreate the same visual experience for the user as if viewing the original scene.2 The tone mapping algorithm uses characteristic of the scene such as contrast, brightness, detail, HVS etc. to approximate the real scene on a display with limited dynamic range. Tone mapping is a well understood problem in literature and works by Eilertsen et al.,3 Aydin et al.,4 and Boitard et al.5 6 are some of the state-of-the-art methods in video tone mapping. On the other hand, tone expansion operators attempt to recreate a scene designed for a SDR display on an HDR display. Methods by Akyuz et al al.7 and Bist et al.8 are well known video tone expansion techniques found in literature. In the case of tone reproduction between HDR content and HDR displays both with different dynamic range we look for common ground in existing methods in tone compression and expansion. It must be noted that we are not compressing or expanding high dynamic range ratios seen in previous studies but dealing with medium dynamic range compression or expansion. Reinhard et al.9 first anticipated the current day scenario with compatibility issues between HDR displays. This work suggested a tone reproduction framework consisting of a forward and reverse transform as seen in figure 1 similar to the ones used in color appearance models.10 The study by Reinhard et al.9 concluded that applying a sigmoid in the forward and backward step amounts to a gamma correction in the form: Ld = c × Ln/m

(1)

where Ld is output display luminance, L is input luminance, c, n and m are constant content and display related parameters. The power function is not an unfamiliar operator in HDR imaging. Early tone mappers such as the one by Tumblin and Rushmeier11 uses gamma correction to model the HVS’s non-linearity. It is known that this method is not ideal for high dynamic range compression ratios but is a possible solution for medium dynamic range compression. Furthemore, recent studies in lightness perception also propose a system gamma based tone mapping12 13 . For a long time in the field of tone expansion, work by Akyuz et al.7 suggesting a linear expansion

Tone Reproduction Operator Reverse Transform

Forward Transform

Expansion

Compression

4000 nits HDR Video

HDR Display

Compression/Expansion

1000 nits

Tone Space

Compression/Expansion

HDR Video

1000 nits HDR Display

Compression

Expansion

100 nits

4000 nits

SDR Video

100 nits SDR Display

Figure 1. A Tone Reproduction Operator using a forward and reverse transform.

operator gave the best results. However, recent work by Bist et al.8 improved this work for stylized content also using a gamma correction expansion. Based on this evidence that gamma correction works well in tone expansion and compression, we test tone reproduction using power function based strategies. Such methods are ideal to test in video due the their global nature as we can avoid temporal artifacts and computational complexity due to their frame based nature.14

2.2 Lighting Style Aesthetics To evaluate the tone reproduction of source content amongst HDR displays we characterize artistic intent in terms of lighting style aesthetics. Lighting styles are used in cinematography to establish narrative tone, realize creative intent and create stylized looks.15 The type of lighting style exploits peak brightness and contrast of the content, hence making it an ideal choice to evaluate tone reproduction in terms of style preservation for HDR. In order to have a good variety of HDR content to challenge the tone reproduction techniques we present three main lighting styles:

Contrast

Low Key (LK)

Bright Key (BK)

Medium Key (MK)

Dark Key (DK)

High Key (HK) Luminance

Figure 2. Lighting style aesthetics distinguished by luminance and contrast.

1. Low Key (LK): A style where the image is globally dark but has small regions of diffuse white or specular highlights. Hence, having low brightness but high contrast. This lighting style is used by cinematographers to capture a captures a dark and dramatic narrative.

2. Medium Key (MK): A style with moderate amounts of brightness and contrast. Most TV video is medium key - sports, news, television series, etc. 3. High Key (HK): A style with bright tones with low contrast. It is often used to express a light mood. This style is commonly found in fashion and advertising. We address these 3 classes of lighting style as LK, MK and HK respectively for the remaining of the document. Absence of high quality and diverse HDR content has been a major hurdle in the HDR research community. This is even more true for stylized content. For this paper we use 4000 nits content from the MPEG test set 16 and a few stylized content from sequences from Froehlich et al.17 The 1000 nits content used are extracted from a footage taken during the French stage of the Volvo Ocean Race. This was done under a collaborative project by French Multimedia companies18 19 .

2.3 HDR Ecosystem With advances in HDR video capture and HDR display technologies the HDR video pipeline has finally began maturing in the industry. Recently, the UHD Alliance (UHDA)20 announced specifications for next generation displays with the following specification: Table 2. UHD Alliance Premium Specification

Specification

LCD HDR Displays

OLED HDR Displays

Content Mastering Displays

Resolution

3840x2160

3840x2160

3840x2160

Bit Depth

Minimum 10-bit

Minimum 10-bit

Minimum 10-bit

Color

BT.2020

21

21

BT.202021

BT.2020 22

More than 90% of P3

22

Minimum 100% of P322

Display Reproduction

More than 90% of P3

EOTF

SMPTE ST 208423

SMPTE ST 208423

SMPTE ST 208423

Peak Brightness

More than 1000 nits

More than 540 nits

More than 1000 nits

Black Level

less than 0.05 nits

less than 0.0005 nits

Less than 0.03 nits

Table 2 gives a good indication of the diversity in display technology (LCD and OLED) for consumer TV’s available in the market. It also shows the gap they have with mastering displays. Traditionally, content has been graded in 100 nits SDR and have been targeted for displays with similar dynamic range. The arrival of these new mastering displays (more than 1000 nits) gives an unprecedented artistic space for content providers. This content is then required to adapt to the target display using tone mapping methods. With increasing trends in peak brightness and lower black levels, we can expect future mastering displays with higher dynamic range. This would also mean that over a period of time a given video library would have HDR content graded at 1000 nits or more and legacy video graded at 100 nits. When this content is transmitted to a consumer HDR TV of a given brightness (540 nits OLED or 1000 nits LCD), the display compresses or expands the dynamic range of the content to exploit the capabilities of the display. This eventuality distorts the content to the extent that the viewer experiences the content differently for a given HDR display. This is a major issue especially when preserving artistic intent of source material and hence underlines the need of good tone reproduction. For experiments in this paper, we have worked with 2 different HDR displays: 1. Sim2 HDR47: 4000 nits prototype LCD display 2. Sony BVM-X300: 1000 nits professional OLED Display The Sim2 display is a prototype HDR display and is commonly used in most HDR studies. The Sony BVM-X300 is an industry compliant professional display with no internal processing. Both the Sim2 and Sony BVM-X300 are ideal displays for our experiment as we have full control over the algorithms applied.

3. TONE REPRODUCTION TECHNIQUES In the following sections we touch upon some basic methods of tone reproduction. First, we experiment with clipping the signal in the case of tone compression from 4000 to 1000 nits content. Second, we have a look into remapping 1000 nits content into a 4000 nits container for tone expansion. Finally, we explore tone reproduction in the form of a power function as promised in section 2.1. All three cases were evaluated with content of various lighting styles.

3.1 Clipping for Tone Compression When working in an HDR post production environment we often observe that content creators don’t necessarily exploit all the high dynamic range available to them. Even when grading on 4000 nit monitors, we identify that artists don’t master indoor scenes in HDR any differently than in SDR. In the case of outdoor scenes, they grade bright whites more brighter than indoor scenes and also preserve a number of specular highlights. However, the major bulk of the scene’s histogram is still SDR. It could be argued that reduction of highlights in the form of clipping can suffice for medium dynamic range compression. From our observation we learn that clipping an image is very much dependent on the type of scene. For a low key scene clipping from 4000 to 1000 nits does make sense since most of the content is within the 1000 nits containers. However, for medium key or high key scenes clipping results in loss of details and sometimes large over saturated regions. Figure 3 show the effect of clipping on different stylized content. Therefore clipping for medium dynamic range compression does not seem like a pliable solution especially for brighter lighting styles.

Figure 3. Clipping for Medium dynamic range compression leads to loss in detail of bright regions. From this figure we see loss in detail: (a) in the flame in low key image (b) along the walls and doors in medium key image and (c) face texture and clothing in high key image. For the high key image, the certain regions seem to be blended with the background. It must be noted that all HDR content in this paper has been tone mapped using display adaptive tone mapping by Mantiuk et al.24

3.2 Remapping for Tone Expansion The UHD forum guidelines25 brings the notion of remapping SDR content into an HDR container. Remapping is the process of repackaging the SDR content in HDR containers without changing the color gamut or the dynamic range of the content. This permits content providers to preserve artistic intent as the pixels are mapped to the equivalent color and luminance values. This also avoids the step of SDR to HDR up-conversion or tone expansion which is not preferred by all artists. We attempted a few tests while remapping 1000 nits content on a 4000 nits SIM 2 monitor. In terms of artistic intent the video remains unchanged. However, when the video is played sequentially in between two 4000 nits videos this results in poor user experience as explained in Figure 4. Hence, it’s likely that remapping may not be the ideal solution to preserve artistic intent as display manufacturers would like to have a constant user experience which requires to adapt the content to tonal range of the display.

Figure 4. Playing videos sequentially from left to right, it is observed that remapping medium or lower dynamic range content along with native HDR content results in a poor user experience.

3.3 Gamma as a Dynamic Range Converter From Section 2.1, we learn that a power function has often been used in tone mapping and tone expansion. Here we explore to use it as a dynamic range converter in the form below: LT arget = α × LM aster

γ

(2)

where LT arget is target luminance and LM aster is mastering luminance, α is maximum target display luminance and γ is the dynamic range conversion value. The target display can be of 4000 nits or 1000 nits and vice versa. Same for the mastering display. The γ value controls this conversion process. Previous methods by Masia et al.26 and Bist et al.8 work towards adaptive gamma correction which is content dependent in tone reproduction. From MPEG, contributions by Lasserre et al.27 also describes the HDR grading process as power function for display referred tone reproduction that is very much scene dependent. This Section looks into gamma correction as a viable tone reproduction operator using lighting style as an indicator of content dependence.

3.4 Experiments For this perceptual study, we aim to answer the following question: At what gamma value does the output video best represent the style of the input video? This study is in two parts. The first part tests tone compression with 4000 nits input video and 1000 nits output. The second examines tone expansion with 1000 nits input video and 4000 nits output. In both studies, the user is requested to choose a gamma value to get a style match. The tests were conducted using ITU-R recommendations BT.500,28 in a dark environment of approximately 25 lux. The content was of LK, MK and HK lighting styles was used. The resolution was 1080p and we tested

with a variety of frame rates up to 24-60 fps. The videos were short in duration about 5-10 seconds. Since we hope to get information about the choice of gamma, we conducted tests by a few video experts with previous experience in the field. We preferred conducting the test sequentially instead of side by side since the brighter display tends derive more attention from the user. For example, in the case of tone compression, we first display the original 4000 nits content on the Sim2 display, followed by a gamma corrected version on the 1000 nits Sony display. After viewing several gamma corrected sequences the user chooses the best gamma value at which they style of the original content was preserved. The user can repeat the test several times for a particular sequence until he/she is satisfied with an appropriate gamma value.

3.5 Results

1

2

0.9

1.8

Gamma (γ)

Gamma (γ)

For tone compression tests, we allow the the user to choose gamma values between 0.5 to 1.2 while for tone expansion the user has the choice between 0.8 and 2.2. These gamma values increment on a step size of 0.1. The results of both experiments can be seen in Figure 5. For tone expansion, we observe that when moving from low key to high key there is a tendency for the gamma value to increase. While for tone expansion, we observe the gamma value decrease as the brightness of the lighting style increases. Both results strongly suggests that for ideal tone reproduction, the gamma value is pretty much style dependent.

0.8

1.6

0.7

1.4

0.6

1.2

0.5

LK

MK Style Key

HK

1

LK

MK Style Key

HK

(b) Tone Expansion (a) Tone Compression Figure 5. Box-and-Whiskers plots per lighting style, displaying gamma preference for (a) tone compression and (b) tone expansion.

One direct observation we can make from these graphs is when the γ choice is 1, which essentially amount to linear scaling. For majority of the content we see that γ = 1 is not chosen. In the case of tone compression, only a few user preferred linear scaling for HK video, while most users preferred a gamma correction. This is because a simple linear scaling quantizes small perceptual differences to the same value on the display resulting is loss of detail. In addition to this, viewers claimed a linear operator is too dim was too dim in our tone mapping studies especially for LK and MK content. For compressing the dynamic range of HK content, majority of the scene is very bright and there is very little dark regions thus the loss of detail caused by linear scaling is relatively less as opposed to LK and MK video. Thus for a HK video, γ ≈ 1, could be acceptable. Similarly, the tone expansion results also shows no preference for linear scaling. Viewers claimed LK and MK content looked too bright and creative intent was clearly lost. This explains the deceasing trend in γ as the brightness of the style increase and is also in accordance with results of Bist et al.8 This is also contradictory to the work by Masia et al.26 which suggests to increase γ value for over-exposed or brighter images. Another interesting result, was regarding the tone expansion of 1000 nits HK content to 4000 nits monitor. There was a tendency amongst users to use higher gamma values in an attempt to reduce the brightness of the scene as many found 4000 nits HK video uncomfortably bright. This could be linked to the same perceptual phenomena that

caused users to choose higher γ in the works by Masia et al.26 Furthermore, we observed that users preferred a more contrasted 4000 nits HK sequence as opposed to style preserving 4000 HK nits video. This tone expansion study also suggests that a brighter video is not always preferable to the user, especially in HK lighting style. For both studies, we observe that for all three styles there are cases where there is an overlap in gamma values suggesting a single gamma value can work on different lighting styles. The main reason for this is lack of number of data points to give statistically plausible results. Lack of data points also suggests that there is very little stylized HDR content available and artist are yet to get comfortable with grading LK and HK content for HDR. Furthermore, the ideal case would be to have a the same content for graded in 1000 nits and 4000 nits to find a gamma correction technique that is invertible. This would mean modeling a single equation that could map content of a given dynamic range on to a display of any dynamic range.

4. CONCLUSION AND FUTURE WORKS Previous works on HDR imaging deals with converting SDR to HDR or HDR to SDR, however there is limited work on tone compatibility between HDR displays. This paper inspires by existing literature and industrial trends in tone mapping and tone expansion and sets out to find plausible tone reproduction techniques between HDR Displays. We take into account lighting style aesthetics as a testing parameter for stylized content. We found that basic methods such as clipping and remapping don’t work as well as gamma correction. We conducted an expert user study using a gamma correction approach and found the choice of gamma strongly depends on the style of the content for best tone reproduction. Through this user study we also saw that linear scaling is not appropriate for tone reproduction and a form of γ correction is often preferred by most users. We summarize our finding in table 3. Table 3. Summary of the Experiments on Tone Compatibility between HDR Displays

Tone Reproduction Operation

Tone Mapping 4000 nits to 1000 nits

Clipping



Remapping

Linear Scaling

Gamma Correction

Tone Compression 1000 nits to 4000 nits Loss of detail in bright regions especially for MK and HK styles

Poor user experience when played sequentially along side Native HDR content



LK and MK content looked brighter and style was not preserved. Also, users found 4000 nits HK content an uncomfortably bright experience. Different Gamma values for different styles in tone reproduction can best preserve lighting style aesthetics. Loss in detail for LK and MK content. LK and MK content looked dimmer and creative intent was lost.

To conclude, tone reproduction between HDR displays is an important issue to be considered for both content providers and display manufacturers in the near future. Our user studies show that best tone reproduction technique could be simply done using a gamma operation. To successfully reproduce artistic intent, we performed a scene by scene user-controlled gamma correction for tone reproduction. Although this method is time consuming and requires extensive man power, more effort is required resolve this problem. For real-time applications, we find that the current state of the art techniques for tone reproduction are not sufficient. For future works, we wish to find real time solution based on modeling a temporally coherent gamma correction curve. This model should preferably be invertible for tone expansion and compression. Furthermore, we plan to extend these tests with same stylized content graded at 4000 nits, 1000 nits and possibly in SDR as well.

ACKNOWLEDGMENTS This work has been achieved within the Institute of Research and Technology b<>com, dedicated to digital technologies. It has been funded by the French government through the National Research Agency (ANR) Investment referenced ANR-A0-AIRT-07.

REFERENCES [1] Reinhard, E., Heidrich, W., Debevec, P., Pattanaik, S., Ward, G., and Myszkowski, K., [High dynamic range imaging: acquisition, display, and image-based lighting ], Morgan Kaufmann (2010). [2] Banterle, F., Artusi, A., Debattista, K., and Chalmers, A., [Advanced High Dynamic Range Imaging: Theory and Practice], AK Peters (CRC Press), Natick, MA, USA (2011). [3] Eilertsen, G., Mantiuk, R. K., and Unger, J., “Real-time noise-aware tone mapping,” ACM Trans. Graph 34(6) (2015). [4] Aydin, T. O., Stefanoski, N., Croci, S., Gross, M., and Smolic, A., “Temporally coherent local tone mapping of hdr video,” ACM Transactions on Graphics (TOG) 33(6), 196 (2014). [5] Boitard, R., Bouatouch, K., Cozot, R., Thoreau, D., and Gruson, A., “Temporal coherency for video tone mapping,” in [SPIE Optical Engineering+ Applications], 84990D–84990D, International Society for Optics and Photonics (2012). [6] Boitard, R., Cozot, R., Thoreau, D., and Bouatouch, K., “Zonal brightness coherency for video tone mapping,” Signal Processing: Image Communication 29(2), 229–246 (2014). [7] Aky¨ uz, A. O., Fleming, R., Riecke, B. E., Reinhard, E., and B¨ ulthoff, H. H., “Do HDR displays support LDR content?: A psychophysical evaluation,” in [ACM Transactions on Graphics (TOG)], 26(3), 38, ACM (2007). [8] Bist, C., Cozot, R., Madec, G., and Ducloux, X., “Style aware tone expansion for hdr displays,” in [Proceedings of Graphics Interface 2016 ], GI ’16, Canadian Information Processing Society, Victoria, B.C., Canada (2016). [9] Reinhard, E., Kunkel, T., Marion, Y., Brouillat, J., Cozot, R., and Bouatouch, K., “Image display algorithms for high and low dynamic range display devices,” Journal of the Society for Information Display 15(12), 997–1014 (2007). [10] Fairchild, M. D., [Color appearance models], John Wiley & Sons (2013). [11] Tumblin, J. and Rushmeier, H., “Tone reproduction for realistic images,” Computer Graphics and Applications, IEEE 13(6), 42–48 (1993). [12] Kane, D. and Bertalm´ıo, M., “System gamma as a function of image-and monitor-dynamic range,” Journal of vision 16(6), 4–4 (2016). [13] Cyriac, P., Bertalmio, M., Kane, D., and Vazquez-Corral, J., “A tone mapping operator based on neural and psychophysical models of visual perception,” in [SPIE/IS&T Electronic Imaging ], 93941I–93941I, International Society for Optics and Photonics (2015). [14] De Simone, F., Valenzise, G., Lauga, P., Dufaux, F., and Banterle, F., “Dynamic range expansion of video sequences: A subjective quality assessment study,” in [Signal and Information Processing (GlobalSIP), 2014 IEEE Global Conference on], 1063–1067, IEEE (2014). [15] Wisslar, V., [Illuminated Pixels: The Why, What, and How of Digital Lighting], Cengage Learning (2013). [16] Touze, D. and Francois, E., “ISO/IEC JCT1/SC29/WG11-m35477: Description of new version of HDR class A and A sequences,” (Feb. 2015). [17] Froehlich, J., Grandinetti, S., Eberhardt, B., Walter, S., Schilling, A., and Brendel, H., “Creating cinematic wide gamut hdr-video for the evaluation of tone mapping operators and hdr-displays,” Proc. SPIE (2014). [18] “Premiere demonstration d’une chaine de television UltraHD HDR (High Dynamic Range) a l’occasion de l’etape francaise de la Volvo Race avec les acteurs bretons de la filiere numerique.” http://www.images-et-reseaux.com/fr/press-releases/ premiere-demonstration-dune-chaine-de-television-ultrahd-hdr-high-dynamic-range (2015). [19] Biatek, T., Hamidouche, W., Travers, J.-F., and Deforges, O., “Optimal Bitrate Allocation for High Dynamic Range and Wide Color Gamut Services Deployment using SHVC,” Data Compression Conference (DCC) (2016). [20] UHD-Alliance, “UHDA Specification,” (2016). [21] ITU-R, “Recommendation BT.2020-1: Parameters Values of Ultra-High Definition Television Systems for Production and International Programme Exchange.” [22] SMPTE, “D-Cinema Quality Reference Projector and Environment for the Display of DCDM in Review Rooms and Theaters,” (2011).

[23] SMPTE, “ST 2084:2014 High Dynamic Range Electro-Optical Transfer Function of Mastering Reference Displays,” (2014). [24] Mantiuk, R., Daly, S., and Kerofsky, L., “Display adaptive tone mapping,” ACM Trans. Graph. 27, 68:1– 68:10 (Aug. 2008). [25] UHD-Forum, “Ultra HD Forum Draft: Ultra HD Forum Guidelines,” (2016). [26] Masia, B., Agustin, S., Fleming, R. W., Sorkine, O., and Gutierrez, D., “Evaluation of reverse tone mapping through varying exposure conditions,” ACM Trans. Graph. 28, 160:1–160:8 (Dec. 2009). [27] Lasserre, S. and Francois, E., “ISO/IEC JCT1/SC29/WG11-m37463: Comments on scene, display reference and HLG (Arib67),” (Feb. 2015). [28] ITU-R, “Methodology for the subjective assessment of the quality of television pictures,” (2012).