Abstract High Dynamic Range （HDR） and Wider Colour Gamut （WCG） content represents a greater range of luminance levels and a more complete reproduction of colours found in real?world scenes. The current video distribution environments deliver Standard Dynamic Range （SDR） signal Y′CbCr. For HDR and WCG content， it is desirable to examine if such signal format still works well for compression， and to know if the overall system performance can be further improved by exploring different signal formats. In this paper， ITP （ICTCP） colour space is presented. The paper concentrates on examining the two aspects of ITP colour space： 1） ITP characteristics in terms of signal quantization at a given bit depth； 2） ITP compression performance. The analysis and simulation results show that ITP 10 bit has better properties than Y′CbCr?PQ 10bit in colour quantization， constant luminance， hue property and chroma subsampling， and it also has good compression efficiency. Therefore it is desirable to adopt ITP colour space as a new signal format for HDR/WCG video compression.
　　Keywords HDR； WCG； Y′CbCr； ITP； ICTCP
　　1 Introduction
　　Current video distribution environments deliver a Standard Dynamic Range （SDR） signal. For SDR content， the common practice is to apply compression on a non?constant luminance （NCL） Y′CbCr colour difference signal defined in ITU?R BT.709 [1] using a gamma transfer function （ITU?R BT.1886 [2]） and non?constant luminance 4：2：0 chroma subsampling. With the advance of display technologies， commercial interests in High Dynamic Range （HDR） and Wide Colour Gamut （WCG） content distribution are growing rapidly. Compared with conventional SDR content， HDR/WCG video content has a greater range of luminance levels and colours found in real?world scenes， and this creates a more pleasant， immersive viewing experience for people with advanced HDR displays. In order to deliver the HDR/WCG content， an HDR/WCG video distribution workflow has to be employed from content creation to final display， which comprises of post?production， encoding， transmission， decoding， colour volumn mapping and display. It is desirable to have a signal format for HDR and WCG content that is not only suitable for efficient image signal encoding， but also suitable for video compression and colour volume mapping. Therefore， for HDR/WCG content distribution， we can examine whether the conventional Y′CbCr 4：2：0 NCL signal format is still a good format to represent HDR/WCG video， and if it still compressed well by a video codec developed using SDR content. In this paper， the main focus is on compression related part （the encoding and decoding blocks） in the distribution pipeline. 　　MPEG is the working group formed by ISO and IEC to create standards for video and audio compression and transmission. In July 2013， MPEG started to look into the problem at the request of several studios and consumer electronics companies [3]. An Ad?Hoc Group （AhG） on HDR and WCG was established to investigate if any changes to the state?of?the?art High Efficiency Video Coding （HEVC） standard [4] are needed for HDR/ WCG video compression. For applications such as Ultra HD Blu?ray disk， it is mandatory that HDR content is transmitted using the HEVC Main 10 profile with metadata in VUI and SEI message [5]. This is commonly referred to as the “HDR?10” solution [6]. The HDR?10 essential metadata includes the signaling of the following video characteristics： SMPTE ST 2084 HDR Perceptual Quantizer transfer function （PQ?TF）， ITU?R BT. 2020 [7] colour primary， and Y′CbCr non?constant luminance in ITU?R BT. 2020. In this paper， the signal format in HDR?10 is referred to as Y′CbCr?PQ 10bit. Other than defining some metadata to represent different video characteristics， “HDR?10” follows closely the common practice of SDR distribution which uses the gamma transfer function， ITU?R BT.709 colour primary， and Y′CbCr non?constant luminance in BT.709.
　　After several rounds of tests and demonstrations， the MPEG HDR/WCG AhG concluded that， for applications that use high bitrate compression， such as the Blu?Ray application， the performance of “HDR?10” seems to be sufficient. For applications that need compression at lower bitrates， such as broadcast and over?the?top （OTT） applications， several shortcomings of “HDR?10” were discovered， suggesting that further improvement might be necessary. In February 2015， MPEG issued a Call for Evidence （CfE） [8] to look for solutions that improve the HDR/WCG video coding performance over HEVC Main 10. A set of anchors targeting broadcast/OTT bitrates are provided using the HEVC Main 10 codec with ST 2084 [9] and ST 2086 support [10]. Anchors generated this way closely model the “HDR?10” distribution system.
　　As active participants of the MPEG committee work on HDR/WCG coding， Arris， Dolby， and InterDigital submitted a joint proposal （the ADI solution） [11] in response to the CfE. The joint proposal provides evidence that with a few new technologies： 1） ITP （ICTCP） colour space； 2） colour enhancement filter； and 3） adaptive reshaping and transfer function， the coding performance can be further improved for HDR/WCG content [12]. 　　This paper mainly focuses on the ITP colour space. The paper tries to answer two questions： 1） what is ITP？ Compared to Y′CbCr?PQ， what is advantage of using ITP？ 2） How does ITP work for compression？
　　The paper is organized as follows. Section 2 describes ITP colour space. Section 2.1 describes ITP conversion workflow. Section 2.2 presents ITP properties. Section 3 presents ITP compression performance for HDR and WCG video compression followed by conclusion in Section 4.
　　2 IPT Colour Space and ITP （ICTCP） Colour
　　Space
　　Non?Constant Luminance （NCL） Y′CbCr is the most frequently used colour space for the distribution of SDR signals. Y′CbCr is a colour difference model derived from nonlinear R′G′B′ signals. For HDR signals， the ST. 2084 （also known as PQ） transfer function is applied in linear RGB space. NCL Y′CbCr has some limitations： 1） it cannot fully de?correlate intensity information from chroma information； 2） it is constrained by RGB colour primaries； therefore， the 3x3 matrix coefficients keep on changing according to RGB colour primaries； 3） its colour difference weights are not based on perceptual model but are derived by filling a colour volume. Constant Luminance （CL） Y′CbCr was added in ITU?R BT. 2020 to adjust Y′CbCr to better de?correlate intensity from chroma. However， the conversion is significantly more complex， making it harder to use in real applications. The IPT colour space is an alternative colour space to de?correlate intensity and chroma information better matching the perceptual mechanisms of the Human Visual System （HVS） [13]. This alternate colour space is more advantageous for HDR and WCG video than the NCL Y′CbCr in signal representation. The advantage of using HVS to derive such a colour space is that the distortion introduced is perceptually minimized.
　　2.1 ITP Conversion Flow
　　IPT uses a colour opponent model which has similar conversion flow to NCL Y′CbCr but more closely mimics the HVS. I corresponds to brightness of the pixel nonlinearly encoded （similar to how the Y′ is encoded in Y′CbCr）， T corresponds to blue?yellow colour perception and P corresponds to red?green colour perception. IPT was first introduced in 1998 and was optimized using a limited set of training data with standard dynamic range and BT. 709 colour gamuts due to the lack of HDR/WCG content at that time [14]. The proposed ITP colour space improves the original IPT by exploring higher dynamic range （up to 10000 nits） and larger colour gamuts （BT. 2020）. Considering the non?trivial changes over the original IPT and to follow the Y′CbCr practice to have blue?related colour component prior to red?related colour component， the new name ITP or ICTCP colour space is adopted to refer to this variation of the original IPT colour space. 　　To better understand ITP， the early stages of human colour vision are described as follows （Fig. 1） [13]：
　　1） Incoming light strikes the three photo receptors （cones） in the eye that have their peak sensitivity in the （L）ong， （M）edium， and （S）hort wavelengths；
　　2） This linear light is transduced （converted） into a non?linear signal response to reduce dynamic range；
　　3） The non?linear output goes through a colour differencing process to extract important information and separates the signal into three distinct pathways；
　　4） The brain sees three colour opponent channels.
　　ITP conversion steps （Fig. 2） are as follows：
　　1） Compute LMS response；
　　2） Apply non?linear encoding PQ；
　　3） Apply colour differencing equation.
　　The similar complexity of the conversion as Y′CbCr allows mass deployment in a wide range of devices. Essentially the existing devices can just change the 3x3 matrix.
　　The conversion matrix from the CIE XYZ tri?stimulus values to LMS （derived RGB2020 to LMS） and LMS to ITP are listed below. Equ. （1） shows the Conversion from XYZ to LMS colour space， （2） shows the Conversion from RGB2020 to LMS colour space， and （3） shows the Conversion from L′M′S′ to ITP colour space. Note that the coefficients in the conversion matrices shown in this paper are the rounded decimal representation of the real conversion matrices. They may have higher precision or fixed point representation depending on implementation needs.
　　[LMS=0.35920.6976-0.0358-0.19221.10040.07550.00700.07490.8434XYZ] （1）
　　[LMS=0.41200.52390.06410.16670.72040.11290.02410.07550.9004RGB] （2）
　　[IPT=0.50000.50000.00001.6137-3.32341.70974.3780-4.2455-0.1325L'M'S'] （3）
　　2.2 ITP Properties
　　When designing a colour space， the main goal is to minimize colour distortion and prevent visible quantization artifacts when images are represented with a given number of digital codewords （i.e.， given bit depth）. Another requirement is to de?correlate the chroma information from luma information to enable colour subsampling， which is important for video compression. In the context of HDR and WCG， and due to various displays in market which supports different dynamic range and colour gamut， a colour space should fit for colour volume mapping as well. In the following， a set of psychophysical experiments have been conducted to validate the advantages of ITP over Y′CbCr?PQ for HDR and wide?gamut imaging. Compared with Y′CbCr?PQ， ITP has the following properties： 　　1） Better signal representation （smaller just?noticeable?difference） in colour quantization in 10 bits；
　　2） Improved intensity prediction （constant luminance）；
　　3） Better predicted lines of constant hue for worst case；
　　4） Friendliness to 4：2：0 chroma downsampling.
　　2.2.1 Baseband Property in Colour Quantization
　　Baseband signal encoding refers to the representation of a linear light HDR signal in integer codewords in a given bit depth. Ideally， the higher the bit depth， the easier the quantized signal can preserve the dynamics in the original linear light signal. However， due to practical considerations， the highest existing pipeline for broadcast is limited to 10 bits. So it is very important to investigate how good ITP 10 bit baseband property can be. It decides how good the signal can start with and thus impact the full chain performance of HDR and WCG content distribution. Mathematical computation shows that ITP has the best overall baseband performance compared with Y′CbCr and Yu′v′ when quantized to 10 bits （Fig. 3）. The industry accepted DE2000 metric is used to measure the visual difference. If the value is below the detection threshold of one “just noticeable difference” JND， no noticeable colour quantization artifact can be observed. The value of dE2000 for Y′CbCr?PQ 10b is between 3.0 and 5.5. For Yu′v′?PQ， it is 2.3. For ITP， it is about 1.0 above 100 nits which is the JND threshold. The better colour quantization property of ITP is due to the fact that ITP is more perceptually uniform than the other colour spaces [15].
　　2.2.2 Constant Luminance Property
　　Constant luminance encoding is more effective in reducing crosstalk between luma and chroma components than the conventional NCL encoding method. Therefore， a colour space which has better constant luminance property tends to have better chroma downsampling， such as 4：2：0. Both subjective experiment and theory shows that ITP outperforms NCL Y′CbCr in intensity prediction. In the subjective experiment， 11 participants matched the intensity of a colour patch with a reference neutral. The data gathered was used to test various colour spaces. ITP outperforms NCL Y′CbCr in intensity prediction （indicated by the higher correlation to the reference） （Fig. 4）. The property is also validated in scientific analysis， where uniformly distributed RGB samples are generated and is converted in ITP， NCL Y′CbCr and CL Y′CbCr， and the correspondence is shown in Fig. 5. The I in ITP correlates better with constant luminance Y′ than the NCL Y′ from Y′CbCr. 　　2.2.3 Hue Property
　　For colour volume mapping， since observers perceiving changes in hue is more impactful than changes in lightness or chroma， it is desirable to have a colour space as hue?linear as possible. Linear hue lines make it very easy to model the mapping process with the hue?preservation requirement. ITP was designed for linear hue property， so it has better hue linearity. A psychophysical experiment was conducted to determine lines of constant hue at multiple hue angles （Fig. 6）. ITP more closely follows the lines of predicted constant hue than Y′CbCr for the worst case measured by the maximum absolute hue deviation. The most notable improvement with hue linearity is the lack of large deviations in ITP as opposed to those found on the right of the constant hue Y′CbCr plot. Table 1 showed the average and maximum absolute hue deviation of ITP and Y′CbCr， respectively.
　　2.2.4 Chroma Downsampling Property
　　Similar to Y′CbCr， ITP is also friendly to 4：2：0 chroma downsampling， which is the common chroma sampling format used in video compression. Table 2 shows the difference in dB in objective metrics computed from a conversion only workflow for ITP and Y′CbCr?PQ. The details of those objective metrics are referred to in [12]. The MPEG CfE chroma down/up sampling filters are applied. The conversion flow is as follows： RGB 4：4：4 （12 or 16 bit depending on content） ?>Y′CbCr?PQ 4：2：0 10 bit /ITP 4：2：0 10 bit ?> RGB 4：4：4 （original bitdepth）. ITP has overall higher PSNR in luminance channel （Y） and overall colour metrics （DE） than Y′CbCr NCL， and is suitable for compression in the 4：2：0 domain.
　　3 ITP Compression Performance
　　As well as having better baseband signal properties than the NCL Y′CbCr?PQ， ITP can compress well. To evaluate the compression performance of ITP， two studies have been conducted： 1） compare transform coding gain with KLT； 2） compare covariance of ITP with Y′CbCr. We formed a test set of 19 frames representing all the scenes from MPEG HDR/WCG AhG sequences， i.e.， one representative frame from each scene.
　　Transform coding gain is one metric to measure compression performance. It is defined by the ratio of the arithmetic mean to the geometric mean of the variances of the variables in the new coordinates， properly scaled by the norms of the synthesis basis functions for nonunitary transforms [16]. The coding gain is usually measured in decibels， representing the reduction in quantization noise by quantizing in the transformed domain instead of the original domain [16]. The test shows that the coding gain of ITP is 12.42 dB， while the coding gain using optimal KLT is 13.16 dB. 　　Another important statistical indicator for compression is the covariance matrix of the signal for 3 channels. We computed the covariance matrix of the test set with BT. 2020 Y′CbCr?PQ and ITP， both in 10?bit 4：4：4 Standard Range.
　　The covariance of 3 channels for Y′CbCr?PQ 10bit case is shown in Table 3， and that for ITP 10bit case is shown in Table 4.
　　By comparing the above two covariance matrix， we found that the variance of P channel is about four times of Cb channel and the variance of T channel is about four times of Cr channel， respectively. The cross?variances of IP and IT are about twice that of YCb and YCr， respectively. The cross?variance of PT is about four times of CbCr. This indicates that if we reduce the signal of P/T by half， i.e.， representing P/T with 9 bit， the covariance matrix should be close to Y′CbCr case. Table 5 is the covariance matrix for the newly generated ITP signal. They are indeed very close to Y′CbCr case in terms of covariance.
　　The conversion only results for I 10bit and PT 9bit compared with Y′CbCr 10bit case is listed in Table 6. By comparing Table 6 with Table 2， the conversion?only benefit of ITP over Y′CbCr is reduced but is still retained for majority testing content. One exception is the sequence BalloonFestival featuring very saturated colours， suggesting 9 bit is not good enough to signal chroma components for this content.
　　The compression simulation was also performed to test the performance of ITP. ITP colour space is implemented in HDRTools 0.8.1 [17]. HM16.2 [18] is used for compression test. The test sequences and targeted bitrate is listed in Table 7. For I 10bit and TP 9bit case， we used the same fixed QP as in the Y′CbCr?PQ anchor case. The compression results showed similar bit rate as the anchor. The BD?rate calculated using the suggested metrics is shown in Table 8. This is a fair comparison with Y′CbCr because a fixed scalar is used for P and T. Alternatively， we can simply encode the signal with HEVC by setting luma bit depth to 10 and chroma bit depth to 9. This shows that for DE100， which is considered as a performance indicator for colour reproduction， ITP with static reshaping can gain 19% over Y′CbCr?PQ.
　　These findings suggest that ITP 10 bit signal contains more colour information than Y′CbCr?PQ 10 bit signal. Since the baseband signal has much better representation in colour， it gives compression much more flexibility for having a “better” signal to start with. Considering this aspect， a technology called adaptive reshaping is incorporated into ITP to adaptively adjust the quantization of luma and chroma components and maximize coding efficiency. The evidence is shown in the MPEG CfE ADI proposal [11] and CfE test results report [19]， and MPEG Core Experiment CE2.1.1 results [20] where advanced reshaping is applied in ITP colour space. In all those tests， ITP has shown superior compression performance compared to the MPEG CfE anchor. Table 9 list results in MPEG HDR/WCG CE2.1.1. 　　When compared on an HDR display， the ITP with advanced reshaping can significantly improve compression performance of the HEVC Main 10 Anchors. The colour patches/blotches are mitigated substantially in low to medium bitrate compression. Besides， it can also improve texture preservation. Fig. 7 shows the snapshots taken during the side?by?side （SbS） viewing on the HDR reference display Pulsar， for the test sequences “Market3” （copyright @ Technicolour） in MPEG CfE ADI solution. The circled areas show the most significant improvements over the anchor. For example， at similar bitrates， the details on the wall and wood frames in Market3 are better preserved （Fig. 7）.
　　4 Conclusions
　　In this paper， ITP 10 bit is shown to have better baseband properties than Y′CbCr?PQ 10 bit. The compression performance of ITP 10 bit is also justified with compression results both in MPEG HDR/WCG CfE and following Core Experiments. The other property of ITP also shows that it is a good fit for colour volume mapping too. ITP is shown to work well for full HDR and WCG video delivery pipeline. Therefore， it is desirable to endorse ITP as a new signal format for HDR/WCG signal.
　　Acknowledgment
　　We would like to acknowledge Y. He， Y. Ye， and L. Kerofsky from InterDigital and D. Baylon， Z. Gu， A. Luthra， K. Minoo from Arris to work together for the joint CfE contribution.
　　References
　　[1] Parameter Values for the HDTV Standards for Production and International Programme Exchange， ITU?R BT.709， 2015.
　　[2] Reference Electro?Optical Transfer Function for Flat Panel Displays Used in HDTV Studio Production， ITU?R BT.1886， 2011.
　　[3] H. Basse， W. Aylsworth， S. Stephens， et al.， “Proposed standardization of XYZ image，” Doc. m30167， Vienna， Austria， Jul. 2013.
　　[4] HEVC， Recommendation ITU?T H.265， International Standard ISO/IEC 23008?2， 2013.
　　[5] Blu?ray Disc Association， “Ultra HD Blu?ray Video Parameters Liaison Information，” Doc. m36740， Warsaw， Poland， Jun. 2015.
　　[6] D. Le Gall， A. Tourapis， M. Raulet， et al.， “High dynamic range with HEVC main10，” JCTVC?U0045， Warsaw， Poland， Jun. 2015.
　　[7] Parameter Values for Ultra?High Definition Television Systems for Production and International Programme Exchange， ITU?R BT.2020， 2015.
　　[8] A. Luthra， E. Francois， and W. Husak， “Call for evidence （CfE） for HDR and WCG video coding，” Doc. N15083， Geneva， Switzerland， Feb. 2015.
　　[9] Electro?Optical Transfer Function for High Dynamic Range Reference Display， Society of Motion Picture and Television Engineers ST 2084， 2014. 　　[10] Electro? Mastering Display Colour Volume Metadata Supporting High Luminance and Wide Colour Gamut Images， Society of Motion Picture and Television Engineers ST 2086， 2014.
　　[11] D. Baylon， Z. Gu， A. Luthra， et al.， “Response to call for evidence for HDR and WCG video coding： Arris， Dolby and InterDigital，” Doc. m36264， Warsaw， Poland， Jul. 2015.
　　[12] A. Luthra， E. Francois， and W. Husak， “Call for Evidence （CfE） for HDR and WCG Video Coding，” Doc. N15083， Geneva， Switzerland， Feb. 2015.
　　[13] G. M. Johnson， X. Song， E. D. Montag， and M. D. Fairchild. “Derivation of a colour space for image colour difference measurement，” Colour Research & Application， vol. 35， no. 6， pp. 387-400， 2010.
　　[14] F. Ebner and M. D. Fairchild. “Development and testing of a colour space （IPT） with improved hue uniformity，” in Colour and Imaging Conference， Scottsdale， USA， Nov. 1998， pp. 8-13.
　　[15] J. Froehlich， T. Kunkel， R. Atkins， et al.， “Encoding colour difference signals for high dynamic range and wide gamut imagery，” in Colour and Imaging Conference， Darmstadt， Germany， Oct. 2015， pp. 240-247.
　　[16] H. Malvar， G. Sullivan， and S. Srinivasan， “Lifting?based reversible colour transformations for image compression”， in Proc. SPIE 7073， Application of Digital Image Processing XXXI， San Diego， USA， 2008. doi： 10.1117/12.797091.
　　[17] MPEG SVN [Online]. Available： http：//wg11.sc29.org/svn/repos/Explorations/XYZ/HDRTools/tags/0.8.1
　　[18] HEVC [Online]. Available： https：//hevc.hhi.fraunhofer.de/svn/svn_HEVCSoftware/tags/HM?16.2
　　[19] MPEG requirement， “Test Results of Call for Evidence （CfE） for HDR and WCG Video Coding”， Doc. N15350， July 2015， Warsaw， Poland.
　　[20] D. Baylon， Z. Gu， A. Luthra， et al.， “CE2.1.1： single layer HDR?only solution based on m36264”， Doc. m37070， Geneva， Switzerland， Oct. 2015.
　　Manuscript received： 2015?11?28
　　Biographies
　　Taoran Lu （tlu@dolby.com） is currently a Staff Researcher with Dolby Laboratories， Inc. She got her PhD in electrical and computer engineering from the University of Florida in December 2010 and joined Dolby in January 2011. Her research interest is on image/video processing and compression， especially on high dynamic range video distribution. She has been an active participant in the MPEG/ITU?T video compression standardization expert group. She has authored and co?authored many academic journal and conference papers， standard contributions and US patents. 　　Fangjun Pu （Fangjun.Pu@dolby.com） is currently a Research Engineer at Image Technology Department in Dolby Laboratories， Inc. She got her Master degree in Electrical Engineering Department from University of Southern California in May 2014. Her research interest is about image/video processing and compression. She is an active participant in MPEG video compression HDR/WCG related standardizations. She has authored or co?authored several standard contributions and conference papers.
　　Peng Yin （pyin@dolby.com） is Senior Staff Researcher at Image Technology Department at Dolby Laboratories， Inc. from 2010. Before joining Dolby Laboratories， she worked at Corporate Research， Thomson Inc./Technicolor. She received her PhD degree from Princeton University in 2002. Her research interest is video processing and compression. She is very active in MPEG/VCEG video coding related standardizations and has many publications and patents. She received the IEEE Circuits and Systems Society Best Paper Award in 2003.
　　Tao Chen （tchen@dolby.com） holds a PhD degree in computer science. Since 2011， he has been with Dolby Labs as Director of Applied Research. Prior to that， he was with Panasonic Hollywood Lab in Universal City， CA and Sarnoff Corporation in Princeton， NJ. His research interests include image and video compression， 3D video processing and system， and HDR video technology. Dr. Chen has served as session chairs and has been on technical committees for a number of international conferences. He was appointed vice chair of a technical group for video codec evaluation in the Blu?ray Disc Association in 2009. Dr. Chen was a recipient of an Emmy Engineering Award in 2008. He received Silver Awards from the Panasonic Technology Symposium in 2004 and 2009. In 2002， he received the Most Outstanding Ph.D. Thesis Award from the Computer Science Association of Australia and a Mollie Holman Doctoral Medal from Monash University.
　　Walt Husak （WJH@dolby.com） is the Director of Image Technologies at Dolby Labs， Inc. He began his television career at the Advanced Television Test Center （ATTC） in 1990 carrying out video objective measurements and RF multipath testing of HDTV systems proposed for the ATSC standard. Joining Dolby in 2000， Walt has spent his early years studying and reporting on advanced compression systems for Digital Cinema， Digital Television， and Blu?ray. He has managed or executed visual quality tests for DCI， ATSC， Dolby， and MPEG. He is now a member of the CTO's office focusing his efforts on High Dynamic Range for Digital Cinema and Digital Television. Walt has authored numerous articles and papers for a number of major industry publications. Walt is an active member of SMPTE， MPEG， JPEG， ITU?T， and SPIE. 　　Jaclyn Pytlarz （Jaclyn.Pytlarz@dolby.com） holds a BS degree in Motion Picture Science from Rochester Institute of Technology. She has worked at Dolby Laboratories since 2014 as an Engineer in the Applied Vision Science Group inside Dolby’s Advanced Technology Group. Prior to work at Dolby， she worked at the Academy of Motion Picture Arts and Sciences as an Imaging Science Intern and iCONN Video Production in 2013 and 2012 accordingly. Her main areas of research include vision and color science as it relates to developing technology for high dynamic range and wide color gamut displays as well signal processing for future compatibility.
　　Robin Atkins （Robin.Atkins@dolby.com） has degrees in Electrical Engineering and Engineering Physics. His career in color and imaging science began while designing High Dynamic Range displays at Brightside Technologies. These displays revealed a fascinating host of new challenges in color appearance， which he is now working to address as part of the Applied Vision Science Group at Dolby Labs. His main focus is on building color management systems for mapping High Dynamic Range and Wide Color Gamut content to a wide range of consumer display devices， and solving the question of how to best represent large color volumes for content distribution.
　　Jan Fr?hlich （jfroe@dolby.com） is PhD?Student at the University of Stuttgart. He is currently working on high dynamic range and wide color gamut imaging and gamut mapping. Froehlich contributed to multiple research projects on new acquisition， production and archiving systems for television and cinema and has been involved in a number of technically groundbreaking film projects， such as Europe’s first animated stereoscopic feature film and the HdM?HDR?2014 high dynamic range & wide gamut video dataset. Before starting the PhD he was Technical Director at CinePostproduction GmbH in Germany. He is member of SMPTE， IS&T， SPIE， FKTG， and the German Society of Cinematographers （BVK）.
　　Guan?Ming Su （guanming.su@dolby.com） is with Dolby Labs， Sunnyvale， CA. Prior to this he has been with the R&D Department， Qualcomm， Inc.， San Diego， CA； ESS Technology， Fremont， CA； and Marvell Semiconductor， Inc.， Santa Clara， CA. He is the inventor of 50+ U.S. patents and pending applications. He is the co?author of 3D Visual Communications （John Wiley & Sons， 2013）. He served as an associate editor of Journal of Communications； and Director of review board and R?Letter in IEEE Multimedia Communications Technical Committee. He also serves as the Technical Program Track Co?Chair in ICCCN 2011， Theme Chair in ICME 2013， TPC Co?Chair in ICNC 2013， TPC Chair in ICNC 2014， Demo Chair in SMC 2014， General Chair in ICNC 2015， and Area Co?Chair for Multimedia Applications in ISM 2015. He is the Executive Director of Industrial Governance Board in Asia Pacific Signal and Information Processing Association （APSIPA） since 2014. He is a senior member of IEEE. He obtained his Ph.D. degree from University of Maryland， College Park.