Re: Specifying an API to query the display color volume

How about this:
Start with standardized values, from maybe 40 to 10,000.
(Such as 48, 80, 100, 160, 203, 235, 300, 500, 600, 720, 1000)
If some standard values are close together, fake example 200 and 203, drop one.
Fill in large gaps (> 40%? increment) with intermediate values

These values would be max luminance for small patches.
Some displays can return this value over HDMI, but the value must be mapped to the limited set to prevent fingerprinting.

No ambient info, as:

  *   this would require a measurement device.
  *   Standardized mastering display color volume does not include it

Lars

On 9/8/23, 11:55 PM, "Andrew Somers" <Andy@GeneralTitles.com<mailto:Andy@GeneralTitles.com>> wrote:


EXTERNAL: Use caution when clicking on links or opening attachments.


IMO, missing is 120, which is referenced in ISO standards, and also a commonly accepted “desktop standard” recited in monitor calibration systems (e.g. Xrite).

ISO 3664:2000  recites “80 to 120” as a desired range.

EBU 3320 recites 100, 200, 250 for SDR and 1000, 600, 500 for HDR

Common mobile devices are capable of 700, 1200, even 1600.

Though also, in theory, SDR displays are (intended) to be calibrated so that 20% of peak white is equivalent to the diffuse surround, or that a the SDR peak is equivalent to a perfect diffuse white reflector in the ambient illumination. For similar perceptual intent, the display’s gamma or effective TRC generally is increased when the peak white of the display exceeds the ambient as described.

I mention this as a step-wise approach may be appropriate, with the oddball examples tacked on (or considered approximated?)

Side note on 48
SMPTE 196M specifies open gate i.e. no film, @16 fL (54.8 cd/m²) and 11fL minimum (old lamp)—with print stock in the gate at Dmin, 16fL screen is ~14fL, which is why 14 fL for DCI. Something of interest though is in the use case of cinema, that 14fL is specular highlights. A 90% white card ends up at about 12 fL (41nits). Depending on how cheap the theater is, screen luminance can be as low as 8fL due to aging xenon bulbs.

Should 48 be the start? What about 40? 48 is specular highlights with a new xenon bulb, but as mentioned ~40 (12 fL) is common as the diffuse white object, and when doing basic white titles, they’ll usually be 11 fL to no brighter than 12 fL.


By Steps...

For an array of standardized luminance values, perhaps:

40, and increments of 8 up to 80, then
increments of 10 up to 120 then
increments of 20 up to 200 then  (or perhaps 20 from 80 to 200?)
increments of 50 up to 500 then
increments of 100 up to 1000 then
increments of 500 up to 3000 then
increments of 1000

This leaves 203, 235, 720 as outliers, but which are reasonably equivalent to 200, 250, 700.


The following items may be tangential, but affect gamut volume, so dropping them here only if they might have some relevance, unsure as I just stumbled onto this thread:


Second: SDR vs HDR
Which level is SDR vs HDR? 1200 might seem likely HDR—but a mobile device @ outdoors day, with indirect sunlight, 1200 is just adequate for SDR. Therefore, explicit demarkation of SDR/HDR is needed as it can not be assumed from peak luminance.


Third: Ambient and  Gamma
This then brings up the ambient light level, and the transfer curve (which is needed for context of the display peak in a given ambient condition), needed to define the expected perception.


Transfer curves (gamma):
For SDR, for power-portion of HLG, and for power-only HDR curves, reasonable steps: 1.0 to 2.0 in 0.2 increments, and 2.0 to 3.0 in 0.1 increments.  ...Plus PQ.


Ambient lux: phi levels or stops?

Ambient is certainly relevant to gamut, but is it to this discussion?
If so, then a reasonable set of values could be a defined mid point such as 210 lux, and an increment by multiplying 210 (lux) by phi^n where n is ±12, examples:


Phi^n levels, mid point 210 lux
   n = ±6 results in 12, 19, 31, 50, 80, 130, 210, 340, 550, 890, 1439, 2329, 3768


Full stops: 2^n levels,  mid point  240 lux - probably not granular enough?
   n = ±4  results in 15, 30, 60, 120, 240, 480, 960, 1920, 3840
Half stops: (sqrt 2)^n levels,  mid point  240 lux
   n = ±8  results in 15, 21, 30, 42, 60, 85, 120, 170, 240, 339, 480, 679, 960, 1358, 1920, 2715, 3840
Third stops: (cuberoot 2)^n,  mid point  240 lux
   n = ±10  results in  24, 30, 38, 48, 60, 76, 95, 120, 151, 190, 240, 302, 381, 480, 605, 762, 960, 1210, 1524, 1920, 2419





Andrew Somers
Senior Color Science Researcher
-contact redacted for public list-
[Image removed by sender.]<https://www.myndex.com/perceptex/>


On Sep 7, 2023, at 7:25 PM, Lars Borg <borg@adobe.com<mailto:borg@adobe.com>> wrote:

In the call we discussed enumerated color volumes.
We also need to enumerate max luminance values (to avoid fingerprinting)
I suggest that the list includes the following nominal max values:
48, 80, 100, 160, 203, 235, 300, 500, 600, 720, 1000

Lars

-----Original Message-----
From: Pierre Lemieux <pal@sandflow.com<mailto:pal@sandflow.com> <mailto:pal@sandflow.com>>
Date: Tuesday, September 5, 2023 at 1:59 PM
To: "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: Specifying an API to query the display color volume
Resent-From: <public-colorweb@w3.org<mailto:public-colorweb@w3.org> <mailto:public-colorweb@w3.org>>
Resent-Date: Tuesday, September 5, 2023 at 1:59 PM


EXTERNAL: Use caution when clicking on links or opening attachments.




Good morning/evening,


In preparation for our call tomorrow and per my long overdue homework,
the PR below updates the strawman by adding an API to query the
display color volume:


https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fw3c%2FColorWeb-CG%2Fpull%2F106%2Ffiles%3Fshort_path%3D10f5ace%23diff-10f5ace43778941f0f6fcac87de875cf240b84fd998411026760e5a573e7c3a9&data=05%7C01%7Cborg%40adobe.com%7C1a44147be605407903d608dbae6c250b%7Cfa7b1b5a7b34438794aed2c178decee1%7C0%7C0%7C638295551726506682%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=JH%2Bb3EBA9Z2oUb6UkVqTCILzVpJDZPNKXOvVDcp719o%3D&reserved=0<https://github.com/w3c/ColorWeb-CG/pull/106/files?short_path=10f5ace#diff-10f5ace43778941f0f6fcac87de875cf240b84fd998411026760e5a573e7c3a9><https://nam04.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fw3c%2FColorWeb-CG%2Fpull%2F106%2Ffiles%3Fshort_path%3D10f5ace%23diff-10f5ace43778941f0f6fcac87de875cf240b84fd998411026760e5a573e7c3a9&amp;data=05%7C01%7Cborg%40adobe.com%7C1a44147be605407903d608dbae6c250b%7Cfa7b1b5a7b34438794aed2c178decee1%7C0%7C0%7C638295551726506682%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&amp;sdata=JH%2Bb3EBA9Z2oUb6UkVqTCILzVpJDZPNKXOvVDcp719o%3D&amp;reserved=0<https://github.com/w3c/ColorWeb-CG/pull/106/files?short_path=10f5ace#diff-10f5ace43778941f0f6fcac87de875cf240b84fd998411026760e5a573e7c3a9>>


The proposed revision also clarifies that the platform is not expected
to perform color volume mapping (i.e. change the colors of the image)
if the image color volume is a subset of the display color volume (as
determined by the API above).


Best,


-- Pierre