Thursday, July 23, 2020
5:00pm - 6:00pm Pacific Time (UTC-12:00am Friday)
Please join us for the next meeting of this virtual working group (VWG). Future meeting dates for this month include:
7/30/20 9:30am
Dropbox Paper link for this group:
We will be using the same GoToMeeting url and phone numbers as in previous groups.
You may join via computer/smartphone (preferred) which will allow you to see any presentations or documents that are shared or you can join using a telephone which will be an audio only experience.
Please note that meetings are recorded and transcribed and open to the public. By participating you are agreeing to the ACESCentral Virtual Working Group Participation Guidelines
Audio Only
You can also dial in using your phone.
Dial the closest number to your location and then follow the prompts to enter the access code. United States: +1 (669) 224-3319 Access Code: 241-798-885
We should test both @jedsmith’s original and @nick’s variation in our testing round
Look into near-primary kinks visible in the polar grid (also secondaries)
Invertibility:
What is that quality metric?
@Thomas_Mansencal: we need invertibility - AR/VR/Game Engines etc - needs to be visually decent
What are the cases for invertibility:
AR/VR
Camera Encoding Primaries
Motion Graphics
General ACES archive
Inverse of even fully saturated colors is already not an issue because of max RGB (max distance of 2)
@carolalynn will put a new section on the Dropbox paper for tests - VFX, Grading, and Other, and footage requirements for each - will start a new post to discuss
Where do we stand on the shadow rolloff / exposure invariance - default to 0 on shadow but possible option to increase for inverse?
@jedsmith’s latest version has control of threshold for all components, so we could for example push less on blue and more on red/green
Threshold parameter should be changed to describe size of confidence gamut, i.e.: 1.0-threshold
About the discussion of hue lines (Munsell vs others):
When comparing the research of hue lines, we need to consider the adaptation state: Munsell used a stable (I believe D50) chromatic adaptation and is focused on large scale equidistance (if this is a word).
Other experiments use other adaptation states and other methods of determining equidistance (JNDs or concatenated matching tests).
They cannot be compared directly.
Also quoting Fairchild from an email group discussion:
" … it is generally recognized that scales of color appearance
(large interval changes in attributes) do not linearly relate to small
interval changes near, and slightly above, threshold changes in
perception."
If you fit for JNDs it won’t agree with other metrics and vice versa.
I don’t think that the gaussian spectral distribution series should guide us, especially outside the confidence zone. And inside the confidence zone, they are not terribly bent, besides the red corner. And the red corner shows that the gaussian approach might have flaws.
Also, you would neglect the existence of colour constancy, which straights those lines anyway.
Cusp in the primaries (hexagonal mode):
I believe, the cusp in the primary is caused by the fact that in the primary hues both non-max components are equal, so no middle-component scaling is happening.
I also think that no natural system would be able to perform the hexagonal model. This might explain why we might prefer the initial version visually.
Shadow roll-off:
You should not forget that lots of cameras produce negative (all three channels) values when underexposed, because of the questionable practice of lens cap black subtraction.
We could add a recommended practice to correctly set the flare before applying gamut compression.
While the classical ~450nm range within sRGB / BT.709 shows skew, it is also worth examining ~612nm. Although it can be argued this is small fish compared to the impact of discrete, per-channel lookups.