- From: Craig Revie <Craig.Revie@FFEI.co.uk>
- Date: Fri, 15 Jun 2018 09:00:47 +0000
- To: Lars Borg <borg@adobe.com>, Phil Green <green@colourspace.demon.co.uk>
- CC: Leonard Rosenthol <lrosenth@adobe.com>, Max Derhak <Max.Derhak@onyxgfx.com>, Simon Thompson-NM <Simon.Thompson2@bbc.co.uk>, "public-colorweb@w3.org" <public-colorweb@w3.org>
- Message-ID: <AM0PR08MB3028AE9411ABDB822BAD5B89D97C0@AM0PR08MB3028.eurprd08.prod.outlook.com>
Hi Lars, As always, your comments are helpful. I understand that you (and many others as shown in your diagram) prefer PQ to HLG. Clearly in cases where there is time for careful colour grading of content this is a good solution. It is not so clear how the PQ model can be used for live streaming of content and it seems that this aspect was the primary motivation for HLG. How is this aspect handled in a PQ world? You ask below whether there are any movies broadcast as HLG and that I don't know (I expect so) but some time ago, the BBC prepared the Blue Planet series as HLG. More recently, the coverage of the wedding of Prince Harry was also prepared as an HLG broadcast and the football World Cup will also be available - all from BBC iPlayer for the time being. Best regards, _Craig From: Lars Borg <borg@adobe.com> Sent: 12 June 2018 05:20 To: Craig Revie <Craig.Revie@FFEI.co.uk>; Phil Green <green@colourspace.demon.co.uk> Cc: Leonard Rosenthol <lrosenth@adobe.com>; Max Derhak <Max.Derhak@onyxgfx.com>; Simon Thompson-NM <Simon.Thompson2@bbc.co.uk>; public-colorweb@w3.org Subject: Re: Help with HLG profile Hello Craig, See below. On 6/9/18, 9:12 PM, "Craig Revie" <Craig.Revie@FFEI.co.uk<mailto:Craig.Revie@FFEI.co.uk>> wrote: Hi Lars, Thanks for your reply. First of all on terminology, is it reasonable to describe HLG as encoding relative scene luminance? Not really. I wouldn't. It's an approximation of scene colorimetry just like 709 or 2020. How accurate is it in real production? And what accuracy do you need? Why? I haven't seen any published tests. But here's what we know from 709 production. I've yet to find anyone that sets their camera in reference 709 mode in real production. Typically 709 productions do not use the standard OETF found in the 709 spec. The camera's linearity response is tweaked by the operator for various reasons. Extra suppression of the darks as a means to reduce noise. And the typical knee for highlight rolloff. (Which HLG now includes in a standard way) Maybe there are options for gamut mapping (which is non-linear) We usually do't call these tweaks color grading, but rather camera setup. Either tweak changes the contrast relations from true scene colorimetry. These tweaks are not indicated in metadata (as HLG has no such, and other formats seem to be proprietary or incomplete), so you have no knowledge on how to undo them in post. I am not very familiar with the technical details of the cameras used for broadcast television but my assumption was that the controls allow white balance, aperture control and selection of filters. I also assume that the sensor is linear So far I agree with you. and that there is minimal processing of the signal. If so, the assumption of relative scene luminance (as described in a number of papers) would seem to be at least a reasonable approximation. Are these assumptions incorrect? I doubt it. As I noted above, there are many in-camera processing options. So the signal might be non-linear. I think that for the uses outlined by Simon on this thread, it would be helpful to have a V4 ICC profile even if there are some limitations (which there may well be). Is this something you could provide? It seems to me that the more useful of the two profiles would be the 'HLG scene profile' although depending on the result of our discussion about terminology this may need a different description. I have yet to find a use case for an HLG scene profile. Please explain how you would use it, and what workflow. Primarily I see post production uses for an HLG reference display profile. With this profile I can mix display-referred content across HLG, PQ, 709 media. For example, the colors in a commercial are display-referred and the repurposing to another media should preserve those colors. Scene-referred conversions wouldn't cut it, so I would not use an HLG scene profile for this. On the point about 'grading', in talking to people from the BBC and NHK, I understand that the main reason to develop HLG was that for many use cases there is no opportunity for grading, for example for live broadcasts. Grading, no. But camera matching, yes, also for live broadcasts. It is the rare event where all cameras are of the same make, model, and revision, including lens and lights. In practice, cameras at major sports events include both HDR and 709 cameras, different sensors, multiple brands, different light setups, different codecs. Same sports jersey => very different colors. So some effort is often spent to make cameras produce similar colors on output. Same sports jersey => similar colors. Most likely that matching process undermines true scene colorimetry. BT.2087 shows conversions from 709 to 2020. My study shows that doing this 'scene'-referred (case 2) creates significant color shifts. In practice, I would have thought that PQ would be used for grading of film production and would be converted to HLG at the time of broadcast. Are there movie HLG broadcasts in Europe yet? PQ for movie grading, yes. Although Adobe Premiere is agnostic and lets you grade (display-referred) HLG content as well. But it seems movie distribution is mostly PQ (HDR10), not HLG. Think Netflix, etc. I like this diagram from Yoeri Geutskens <https://www.linkedin.com/in/yoerigeutskens/> More names in the PQ circle than in the HLG circle. [cid:image001.png@01D4048C.EDF07B80] Lars Best regards, _Craig ________________________________ From: Lars Borg <borg@adobe.com<mailto:borg@adobe.com>> Sent: 10 June 2018 04:53:58 To: Craig Revie; Phil Green Cc: Leonard Rosenthol; Max Derhak; Simon Thompson-NM; public-colorweb@w3.org<mailto:public-colorweb@w3.org> Subject: Re: Help with HLG profile Hi Craig, A very sensible question. Rec. 2100 gives you two reference decodings for HLG media: scene referred and display referred @ 1000 nits. Scene versus display is not a metadata item, but rather a choice by the reader. Presumably the camera operator optimized the camera rendering for a pleasing appearance on the reference display, so 'scene' is a misnomer, as it's always been also with Rec. 709. At the consumer end, each HDR TV set (PQ or HLG) is expected to apply some form of re-rendering based on its assumed luminance (assumed as it is not actually measured or calibrated). That would be your metadata source. However, we should not expect that they implement anything like what's indicated in Rec. 2100. Power limits, motion enhancements, user preferences, etc. are out of scope for Rec.2100. This makes it rather meaningless to try to model anything other than the reference display. In a separate expert forum, a majority stated that even the best consumer HDR TV sets are so bad, they should never be used of color grading. So this begs the question, who would need a parameterized HLG decoder? BTW, the HLG scene profile is trivially implementable in V4. The reference display profile is a little bit more complex in V4, but doable. Thanks, Lars Borg | Principal Scientist | Adobe | p. 408.536.2723 | c. 408.391.9479 | borg@adobe.com<mailto:borg@adobe.com> On 6/8/18, 9:46 PM, "Craig Revie" <Craig.Revie@FFEI.co.uk<mailto:Craig.Revie@FFEI.co.uk><mailto:Craig.Revie@FFEI.co.uk><mailto:Craig.Revie@FFEI.co.uk%3e>> wrote: Hi Max and Phil, I hesitate to ask as this may show my ignorance but I had thought that HLG encodes relative scene luminance and does not carry any metadata. In my understanding the choice of the reference white is made by the camera/cameraman shooting the scene and the camera signal is encoded as HLG which makes this model ideal for broadcast television. You seem to anticipate having max and min luminance values - where do these come from? What does the profile represent and how do you anticipate it being used? As I say, this may just show my ignorance... _Craig On 8 Jun 2018, at 23:18, Phil Green <green@colourspace.demon.co.uk<mailto:green@colourspace.demon.co.uk><mailto:green@colourspace.demon.co.uk><mailto:green@colourspace.demon.co.uk%3e>> wrote: In my understanding yes, but you would not be able to pass in the max and min luminance so would need a different profile for each condition supported. Phil On 08/06/2018 16:28, Leonard Rosenthol wrote: Can this profile be defined WITHOUT the calculator? Leonard From: Max Derhak <Max.Derhak@onyxgfx.com<mailto:Max.Derhak@onyxgfx.com>><mailto:Max.Derhak@onyxgfx.com> Date: Wednesday, June 6, 2018 at 2:43 PM To: Simon Thompson-NM <Simon.Thompson2@bbc.co.uk<mailto:Simon.Thompson2@bbc.co.uk>><mailto:Simon.Thompson2@bbc.co.uk>, "public-colorweb@w3.org<mailto:public-colorweb@w3.org>"<mailto:public-colorweb@w3.org> <public-colorweb@w3.org<mailto:public-colorweb@w3.org>><mailto:public-colorweb@w3.org> Subject: RE: Help with HLG profile Resent-From: <public-colorweb@w3.org<mailto:public-colorweb@w3.org>><mailto:public-colorweb@w3.org> Resent-Date: Wednesday, June 6, 2018 at 2:43 PM Simon has indeed helped me, and I'm very grateful for his assistance. As it turns out there was a small bug in my matlab/octave code. Based on my matlab/octave code I have been able to create iccMAX profiles for both narrow range and full range encoding of Rec 2100 using Hgl curves. I created XML representations of the profiles and used the iccFromXML tool from the reference implementation to create the profiles for testing using the iccMAX reference implementation CMM. Significant features of these prototype profiles include: * Full floating point based algorithmic encoding of RGB to XYZ and XYZ to RGB tags using calculator processing elements * Uses a D65 based PCS * Max and min luminances can be passed in as CMM environment variables to adjust Hlg curves * The profiles use display relative Y=100 for max white for default relative intent processing * Scene relative luminances can be supported with CMM luminance matching using the spectral viewing conditions tag (part of Profile Connection Conditions - PCC) illuminant white point luminance. (Note: Using CMM control options for luminance based matching by CMM allows for scene relative luminances to be used and adjusted for. The PCC can be externally substituted to define an alternate white point luminance for luminance scaling, but matching values for CMM environment variables are also needed to ensure that the corresponding Hlg curves are used). An Interoperability Conformance Specification (ICS) is needed to help define specific profile and CMM requirements for this type of profile. That is something that the ICC display working group will work on. None of this can be directly accomplished using a single V4 profile. V4 uses display relative with a D50 PCS (with chromatic adaptation applied). By playing around with defining the media white point one can achieve a level of luminance scaling using absolute intent with some possible interoperability issues to contend with. The Hlg algorithm cannot be directly encoded in v4, and LUTs and inverse LUTs using interpolation need to be populated for a fixed scene luminance condition with different profiles created for different scene luminances. Regards, Max Derhak (PhD) Principal Scientist From: Simon Thompson-NM [mailto:Simon.Thompson2@bbc.co.uk] Sent: Tuesday, May 29, 2018 8:52 AM To: public-colorweb@w3.org<mailto:public-colorweb@w3.org><mailto:public-colorweb@w3.org> Subject: Re: Help with HLG profile Hi Chris, all, I've already sent a code correction to Max, so hopefully that will fix the issue he was seeing. I'll comment further inline, below. From: Chris Lilley [mailto:chris@w3.org] On 17-May-18 23:25, Max Derhak wrote: Hi, I have an action item in the ICC Display Working group to develop an HLG based iccMAX display profile. In order to understand how to go about it I've first prototyped functionality in Matlab/Octave. The attached zip file has my code along with the BT2100 document that I'm using to implement.. The problem is that for a display profile you need to have both device to PCS as well as PCS to device transforms. The device to PCS transform is conceptually implemented in HLG_FullToXYZ.m, and the PCS to device is conceptually implemented in HLG_XYZToFull.m (I'm using full 0.0 to 1.0 range encoding in this case). Alternatively one could use the HLG_Narrow12ToXYZ.m and HLG_XYZToNarrow12.m to use the narrow 12-bit integer encoding. The problem is that the OOTF (implemented in HLG_EOTF.m) and inverse OOTF (implemented in HLG_invEOTF) functions are not logical inverses of each other (from what I can tell from the docs). I recall hearing that these are not round-trippable. The equations for HLG are reversible and we have used the reverse transforms for converting other HDR formats to HLG, converting SDR camera feeds to HLG and converting HLG to SDR. An example use case is given in: https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.bbc.co.uk%2Frd%2Fblog%2F2018-05-ultra-high-definition-dynamic-range-royal-wedding-uhd-hdr&data=02%7C01%7Cborg%40adobe.com%7Cc687232c3775491b340f08d5cea18e1c%7Cfa7b1b5a7b34438794aed2c178decee1%7C0%7C0%7C636642115681164772&sdata=eICRO%2Fa3ISG%2Fxm50j3UznDG5HMD1Ekf5YHc6uCxp5%2F4%3D&reserved=0<https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.bbc.co.uk%2Frd%2Fblog%2F2018-05-ultra-high-definition-dynamic-range-royal-wedding-uhd-hdr&data=02%7C01%7Clrosenth%40adobe.com%7Cb810dc2b06c24735603308d5cbdd749e%7Cfa7b1b5a7b34438794aed2c178decee1%7C0%7C1%7C636639074360711277&sdata=R%2BFSqmXWBTIxF8EfODM8ZR83S4jkiC8YwMpWTlFzi4Q%3D&reserved=0><https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.bbc.co.uk%2Frd%2Fblog%2F2018-05-ultra-high-definition-dynamic-range-royal-wedding-uhd-hdr&data=02%7C01%7Cborg%40adobe.com%7Cc687232c3775491b340f08d5cea18e1c%7Cfa7b1b5a7b34438794aed2c178decee1%7C0%7C0%7C636642115681164772&sdata=eICRO%2Fa3ISG%2Fxm50j3UznDG5HMD1Ekf5YHc6uCxp5%2F4%3D&reserved=0%3chttps://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.bbc.co.uk%2Frd%2Fblog%2F2018-05-ultra-high-definition-dynamic-range-royal-wedding-uhd-hdr&data=02%7C01%7Clrosenth%40adobe.com%7Cb810dc2b06c24735603308d5cbdd749e%7Cfa7b1b5a7b34438794aed2c178decee1%7C0%7C1%7C636639074360711277&sdata=R%2BFSqmXWBTIxF8EfODM8ZR83S4jkiC8YwMpWTlFzi4Q%3D&reserved=0%3e> On the other hand the BBC claim that the transcode looks "identical" A more complete description of the process of transcoding between HDR formats is given in ITU-R BT.2390 section 7: https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.itu.int%2Fdms_pub%2Fitu-r%2Fopb%2Frep%2FR-REP-BT.2390-4-2018-PDF-E.pdf&data=02%7C01%7Cborg%40adobe.com%7Cc687232c3775491b340f08d5cea18e1c%7Cfa7b1b5a7b34438794aed2c178decee1%7C0%7C1%7C636642115681164772&sdata=ZLkH19GPIpEJ3v48wDdMaoIkMDT8aGfgOleSjeXPVoI%3D&reserved=0<https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.itu.int%2Fdms_pub%2Fitu-r%2Fopb%2Frep%2FR-REP-BT.2390-4-2018-PDF-E.pdf&data=02%7C01%7Clrosenth%40adobe.com%7Cb810dc2b06c24735603308d5cbdd749e%7Cfa7b1b5a7b34438794aed2c178decee1%7C0%7C1%7C636639074360711277&sdata=cXEstDUoMa2r2vIP1fmlx57urfV5uyxbR2hOEjEFbQc%3D&reserved=0><https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.itu.int%2Fdms_pub%2Fitu-r%2Fopb%2Frep%2FR-REP-BT.2390-4-2018-PDF-E.pdf&data=02%7C01%7Cborg%40adobe.com%7Cc687232c3775491b340f08d5cea18e1c%7Cfa7b1b5a7b34438794aed2c178decee1%7C0%7C1%7C636642115681164772&sdata=ZLkH19GPIpEJ3v48wDdMaoIkMDT8aGfgOleSjeXPVoI%3D&reserved=0%3chttps://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.itu.int%2Fdms_pub%2Fitu-r%2Fopb%2Frep%2FR-REP-BT.2390-4-2018-PDF-E.pdf&data=02%7C01%7Clrosenth%40adobe.com%7Cb810dc2b06c24735603308d5cbdd749e%7Cfa7b1b5a7b34438794aed2c178decee1%7C0%7C1%7C636639074360711277&sdata=cXEstDUoMa2r2vIP1fmlx57urfV5uyxbR2hOEjEFbQc%3D&reserved=0%3e> *Also it would be incredibly helpful to better understand what is the purpose of having an HLG ICC profile? There are a few general use cases that we've thought of so far which apply equally to all HDR variants: * Storage of subtitles and other still images - currently still formats like PNG can only store a gamma value in the header files - the only way to store non-gamma encoded images is to embed an ICC Profile. Currently for PQ HDR, there's a draft proposal that looks for a given ICC Profile filename in the header which then over-rides both the header and profile. We would prefer to have the correct ICC profiles available. * Allowing operating systems to correctly display images on non-gamma displays. * IP disptribution platforms (both PC based and Set-top box will need to display video and background images) - e.g. a video embedded in a webpage, you may have a video box in HDR and a surrounding webpage in sRGB, both of which need to be correctly displayed and need an ICC profile description. Any transforms from SDR to HDR will need to place the SDR diffuse white at the correct signal level (given in ITU-R BT.2408) - I'm not sure how this is encoded in to the transforms. Best Regards Simon -- Simon Thompson MEng CEng MIET Project R&D Engineer BBC Research and Development South Laboratory ________________________________ [FFEI Limited]<https://na01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fwww.ffei.co.uk&data=02%7C01%7Cborg%40adobe.com%7C95e992930a4b4a0f85c308d5cddd420a%7Cfa7b1b5a7b34438794aed2c178decee1%7C0%7C1%7C636641272535313819&sdata=Q1alNsLsCwIaIzgnNuFnw45OG31IyZE1TqVlXo8XBCM%3D&reserved=0> CONFIDENTIALITY AND DISCLAIMER NOTICE This message and any attachment is confidential and is protected by copyright. If you are not the intended recipient, please email the sender and delete this message and any attachment from your system. Dissemination and or copying of this email is prohibited if you are not the intended recipient. We believe, but do not warrant, that this email and any attachments are virus free. You should take full responsibility for virus checking. No responsibility is accepted by FFEI Ltd for personal emails or emails unconnected with FFEI Limited's business. FFEI Limited is a limited company registered in England and Wales (Registered Number: 3244452). [Join us on Linked In]<https://na01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fwww.linkedin.com%2Fcompany%2Fffei&data=02%7C01%7Cborg%40adobe.com%7C95e992930a4b4a0f85c308d5cddd420a%7Cfa7b1b5a7b34438794aed2c178decee1%7C0%7C0%7C636641272535313819&sdata=e4Mg0RADSEh5xmatm7jdcxGZ9Og0E9bnIj%2BAA5HK2kM%3D&reserved=0>[Follow<https://na01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fwww.linkedin.com%2Fcompany%2Fffei&data=02%7C01%7Cborg%40adobe.com%7C95e992930a4b4a0f85c308d5cddd420a%7Cfa7b1b5a7b34438794aed2c178decee1%7C0%7C0%7C636641272535313819&sdata=e4Mg0RADSEh5xmatm7jdcxGZ9Og0E9bnIj%2BAA5HK2kM%3D&reserved=0%3e%5bFollow> @FFEI_ltd]<https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Ftwitter.com%2FFFEI_ltd&data=02%7C01%7Cborg%40adobe.com%7C95e992930a4b4a0f85c308d5cddd420a%7Cfa7b1b5a7b34438794aed2c178decee1%7C0%7C0%7C636641272535313819&sdata=WAHld0A%2BGPGtnugMfF52J1cfJignDk0J4%2FFrsHGZ%2F34%3D&reserved=0>[FFEI<https://na01.safelinks.protection.outlook.com/?url=https%3A%2F%2Ftwitter.com%2FFFEI_ltd&data=02%7C01%7Cborg%40adobe.com%7C95e992930a4b4a0f85c308d5cddd420a%7Cfa7b1b5a7b34438794aed2c178decee1%7C0%7C0%7C636641272535313819&sdata=WAHld0A%2BGPGtnugMfF52J1cfJignDk0J4%2FFrsHGZ%2F34%3D&reserved=0%3e%5bFFEI> YouTube Channel]<https://na01.safelinks.protection.outlook.com/?url=http%3A%2F%2Fwww.youtube.com%2Fuser%2FFFEIPrintTechnology&data=02%7C01%7Cborg%40adobe.com%7C95e992930a4b4a0f85c308d5cddd420a%7Cfa7b1b5a7b34438794aed2c178decee1%7C0%7C0%7C636641272535313819&sdata=5G6oBnD43AAKg976Cs1M1%2FlXFf0Urol%2BjaQoIDY3CTI%3D&reserved=0> Registered Office: The Cube, Maylands Avenue, Hemel Hempstead, Hertfordshire, HP2 7DF, England.
Attachments
- image/png attachment: image001.png
Received on Friday, 15 June 2018 09:01:22 UTC