W3C

- DRAFT -

High Dynamic Range breakout, TPAC 2015

28 Oct 2015

See also: IRC log

Attendees

Present
wdh, markw, nigel, skim13, olivier
Regrets
Chair
Mark Watson
Scribe
nigel, markw

Contents


<nigel> scribe: nigel

markw: This seems like a topic that hasn't been discussed much in W3C.
... There may be people here who are interested but not understand it.
... Is there anyone here who is not familiar with High Dynamic Range (HDR) and
... Wide Color Gamut (WCG)?

group: [would like the introduction]

markw: Presents slides
... [We've had more resolution and framerate, now looking for 'better pixels']
... HDR is about luminance, WCG is about colour range.
... [slide with london bus] This isn't the real colour of a London Bus.
... [Luminance slide] Luminance is measured in candela / metres squared or nits
... range of perception is much wider than what SDR TV and HDR TV cover.
... HDR is much bigger than SDR - the brights are brighter and the darks are darker and
... more detail can be viewed in the dark and light regions.
... [Slide showing colour gamut] We take 3 points on the colour diagram and all the
... colours we can represent are inside the triangle that they define. Shows Rec709 as
... the smallest triangle, Rec2020 as the biggest, some in between. There aren't many
... displays that can show the whole 2020 colour space. We can signal and encode
... colours in that space but the device will shrink those down. Lasers are needed right now
... to show the whole gamut.
... [Slide showing 3d picture of WCG and HDR] We talk about colour volume now,
... representing the entire range of colours and brightness that we can encode.
... [Slide, example from Dolby with picture of flowers] All of the red points are those that
... fit in the color volume, and the blue ones are outside it. An ordinary picture has
... colours in it that are way outside the range that we can represent today.

Bill: There's a lovely example of why this matters. Two NY Giants jerseys. A 30 year old
... one and a 10 year old one. Part of their brand identity is the exact colour that they
... use. If you're limited to RGB or Rec709 you cannot accurately and reliably capture the
... colour that is part of the brand identity. So there's a commercial gain to doing this
... more accurately.

wdh: Cinema displays can show a bigger volume than TVs. When you go through the
... process of making a film for cinema and TV you force the colour grader to make
... limiting choices.

markw: For video and contribution display the introduction of HDR and WCG affects all
... the steps in the chain. Video cameras can capture a lot. On the production side
... lights need to be considered too. In Post production color grading and reference monitors
... need to be changed. They need to learn how to use the new tools.
... In Encoding and Delivery you need more bit depth (10 bit), a different transfer function
... (that maps from a 10 bit number to luminance value), and colour space.
... In displays there's lots of new technology in consumer displays - LEDs with local
... backlights and dimming, OLEDs etc. "Tone mapping" is important - the fact that a
... signal that you've encoded has been created for a reference master display. No
... consumer displays have the same brightness as the master display. Tone mapping
... is far from straightforward, to map from one to the other.
... [Web Platform Issues]. Questions for the Graphics plane and the Video Plane.
... There will be devices that support HDR for video but not for graphics. In both cases
... we have both an encoding problem and a discovery problem, to discover display
... capabilities. There's also the question of image formats for HDR images.
... There's video capture on the web platform too.

YYY: for webcam if you have strong backlight, e.g. for video conferencing, it is very
... difficult to get a good picture.

wdh: Consumer capture is really interesting.

markw: Codec and display capabilities are two independent things. Our current APIs
... are quite good at codec discovery - MIME types with codecs, levels, profiles etc.
... But just before I have an HDR codec does not mean I have an HDR display.

nigel: There are also questions around compositing graphics on video. Plus there was
... a liaison from ATSC to TTWG asking about how to deal with HDR and WCG given that
... timed text formats only support sRGB.

markw: Is there a requirement around compositing?

wdh: It feels like an implementation thing.

bill: Don't you need to specify an EOTF.

pierre: If you composite SDR over HDR then there are circumstances when you might want
... to boost or darken the graphics compared to the image. Using an 80 nit max for
... sRGB might be too dark or too light depending on the image.

markw: So you want the application to have a brightness control?

pierre: Give the author the ability to specify that.

markw: That's what I meant.

bill: Having a brightness knob on the entire plane might be an interim solution.

pierre: You don't want to have to backport HDR into CSS, SVG, TTML etc. Also if you have
... an existing library of SDR content does it make sense to make an HDR version of the
... same content?

bill: Sometimes yes, sometimes no, depending on the commercial motivation.

nigel: One thing that's really easy to specify is that sRGB values must not be interpreted
... as HDR values, because that would result in full intensity values being painfully bright.
... (accepting that the relative brightness might need to shift up and down).

dsinger: Actually displays are already a bit brighter than sRGB peak white (80 nits).

markw: You might have to set the brightness control right.

wdh: You've identified these are interesting problems. Are there others?

YYY: Do you mean the graphics plane as encrypted?

markw: I'm talking about the display. Televisions have this problem particularly.
... They have two separate pipeline processes for video vs graphics.

bill: What we've seen is that they do some mapping.

dsinger: There's another issue. In the world of audio we've had to manage range to
... avoid hurting people's ears. In the pictures we may need to do something similar.

wdh: In the UK there are requirements about that.

group: discusses the calm act and the need to make content accessible.

wdh: The brightness issue is valid - just dealing with ambient light feels like an implementation
... as opposed to a standards thing.

bill: The tone mapping is outside the standards space.

markw: A possible technical area that might require work is: You have a site with HDR
... graphics and some adverts. Should the adverts be able to constrain the page peak brightness?

bill: It's not just peak brightness - the appearance isn't just defined by the peak.

nigel: Are you talking about dynamic adjustment of the whole composited plane?

markw: I'm talking about content in an iframe changing the appearance of the main content.

dsinger: Like a dark goth page that's spoiled by a bright pink advert.

nigel: Isn't that a content provider issue?

markw: Yes you could request that only adverts conforming to a particular profile are served.
... If the web platform could in some way cap or constrain the brightness of content
... that could be useful.

bill: An argument in favour is the health safety.

markw: That would argue for some kind of transformations in the CSS realm.

wdh: That's interesting.

markw: The immediate thing that's of most concern is video display capability discovery.
... I thought at some point that CSS was the right place to discover this, but in realising
... that the graphics capability and the display capability might be different I think that
... they're separate.

nigel: So if you discovered you had a sRGB display or a 2020 display what would you do?

markw: You'd choose which content to download and show.

nigel: Why wouldn't you apply the transformation locally to the same content?

dsinger: Some high value content is carefully graded to be optimal for a particular display.

wdh: So if you signal a particular codec and you're casting it to the TV, that doesn't
... allow enough information about what exactly to put on the page. Plus the video window
... might be on a monitor or dragged over to a TV.

dsinger: You can ask 'do you have a 10 bit codec' but that doesn't tell you anything about
... the color space. We've traditionally thought of the display capabilities as being
... independent of the codec.

markw: I've even wondered if you might have a situation where you have an HDR display
... but somehow the content that you fetch is SDR.

YYY: You might have a problem where the display falls back to the lowest common denominator space.

markw: You mean if you were compositing graphics on video?

YYY: Yes.

markw: So if we thought that compositing SDR graphics on HDR video would cause the
... whole thing to drop down to SDR we wouldn't do that.
... From our company perspective everything has to all work together and look
... right for the product. A question in this group would be, for the web platform, if
... we are going to introduce HDR, should we place the same requirements on it.
... You're not allowed to advertise HDR capability unless you can successfully composite
... these things together.

wdh: Set top boxes have a hard problem - multiple in, multiple out. It's even harder than
... for TVs. We're doing it but it's ugly.

dsinger: I've seen an example where negative primary values can be used to go outside
... the color gamut.

markw: +1 to just specifying a new color gamut.

dsinger: Practically, a media query in CSS needs to come up with a boolean answer.
... I'm wondering if asking "What %age of the sRGB or 2020 triangle do you claim to be
... able to render?" then maybe that would work given that the triangle is normally
... reasonably well centred, so you could get close enough that way.

pierre: The challenge is that there are lots of other parameters, not just color volume,
... but also max and min luminance. What about querying specific formats from the display.
... rec709, BT2020 etc.

markw: I think that you can assume that if a display can understand the color values in
... 10 bits then it can display them.

<markw> bill: boolean for colorspace is probably ok

<markw> billh: seeing that luminance threshold depends on device : OLED vs LCD etc.

<markw> ... 400nits OLED may be ok, but 700 for LCD

<markw> dsinger: displays will improve over time, unlike the static situation we believed we were in before

markw: We're near the end of the session. We should try to input this to the CSS WG
... as a starting point to generate some more discussions. I don't have any bright ideas
... about indicating the display capability.
... Is that a category - do still images go with the video capability or the graphics capability?

dsinger: Is the TV rendering it all?

pierre: Captions - the television composites them on the video.
... What's the process in CSS WG?

markw: The process is first find a browser who's interested and then take it to the group!

dsinger: I'm going to talk to my guys about this and do some experiments.

pierre: Do you think they'd be interested to specify colours differently or assume that
... CSS will continue to render to sRGB or both?

dsinger: CSS will need to be able to represent colours outside the sRGB plane.

nigel: Then you have the reverse problem - what do you do if a colour outside sRGB
... needs to be rendered on an sRGB display?

dsinger: yes.
... Just to frighten you, what if someone just offers full ICC colour profile?

pierre: ICC has the concept of diffuse white, which PQ doesn't have.

markw: What is diffuse white?

pierre: When you define an ICC colour profile you have to define diffuse white. Today
... most implementations match that to their maximum white luminance. An example
... is a HDR PNG with an embedded ICC profile that specifies diffuse white at 100 nits.
... Most applications just clip everything above 100.

<scribe> ACTION: markw to forward these minutes to the CSS group! [recorded in http://www.w3.org/2015/10/28-hdr-minutes.html#action01]

<scribe> scribe: markw

Summary of Action Items

[NEW] ACTION: markw to forward these minutes to the CSS group! [recorded in http://www.w3.org/2015/10/28-hdr-minutes.html#action01]
 
[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.140 (CVS log)
$Date: 2015/10/28 08:03:48 $

Scribe.perl diagnostic output

[Delete this section before finalizing the minutes.]
This is scribe.perl Revision: 1.140  of Date: 2014-11-06 18:16:30  
Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/

Guessing input format: RRSAgent_Text_Format (score 1.00)

Succeeded: s/XXX/wdh/
Found Scribe: nigel
Inferring ScribeNick: nigel
Found Scribe: markw
Scribes: nigel, markw

WARNING: No "Topic:" lines found.

Present: wdh markw nigel skim13 olivier
Got date from IRC log name: 28 Oct 2015
Guessing minutes URL: http://www.w3.org/2015/10/28-hdr-minutes.html
People with action items: markw

WARNING: No "Topic: ..." lines found!  
Resulting HTML may have an empty (invalid) <ol>...</ol>.

Explanation: "Topic: ..." lines are used to indicate the start of 
new discussion topics or agenda items, such as:
<dbooth> Topic: Review of Amy's report


[End of scribe.perl diagnostic output]