Output Transforms Architecture VWG - April 3rd, 2024

The recording is now available.

Not sure if this is the right place to discuss this but I watched the meeting and naming for srgb vs 2.2 was discussed.

In the broader sense of naming conventions. Is there any specific reason or desire to name them as “full” color spaces/primaries only?

In the logical sense I think you could name one sRGB - sRGB and the other sRGB - Gamma 2.2. I believe in Baselight the former is referred to as ~2.2.

But whichever solution gets picked, wouldn’t these two ODTs now be the only one where naming convention deviates?

What about the desktop monitors that do P3? Apple calls it Display P3 but it could be named P3 - sRGB. What about other modern PC displays that do P3 - Gamma 2.2 instead?

If tidyness or simplicity is the main reason, I personally find seeing more information with less ambiguity a better approach and less confusing. And maybe sRGB - sRGB reads a bit odd but perhaps less when all ODTs note their transfer function after a dash.

2 Likes

I tend to agree, it’s just weighing verbosity vs convenience. Here is the v2 transform list right now, which I have updated slightly after reviewing the names used in OCIO v2.

For the HDR transforms we already include the “ST2084” or “HLG” after the primary encoding, but currently just assume the transfer function and use shorthand for things like “Rec.709” or “sRGB”. However, it probably wouldn’t be the worst thing to just add verbosity and include a a “Gamma 2.6” after all the P3 projection ODTs or a “BT.1886” after the SDR Rec.709/Rec.2020 names so that each has a "Primary TransferFunction (x nit / Dxx sim / Limited)

3 Likes

I think, P3 gamma 2.2 should also be added. In sRGB vs pure gamma thread there was a discussion about Apple P3 display curve. If I recall it right, it’s sRGB in some P3 devices and Gamma 2.2 in others.

I also think, it should be Gamma 2.4 instead of BT1886 in the name. Even colorists sometimes don’t know what BT1886 is, not to mention most of the other artists who use ACES. Also, I’ve only seen one TV, that had an option to set it to BT1886 curve. Most TVs are in Gamma 2.2 by default. And Gamma 2.4 as one of gamma presets.

Even grading on BT1886 calibrated display is a debatable thing. Not sure if @steve from Light Illusion will get the notification, since he is offline for quite a long time. But he for sure could provide good statistics on how often people choose BT1886 vs Gamma 2.4 for calibration.

So, while in case of ODT, BT1886 and Gamma 2.4 both mean the same thing, I think, the better name would be Gamma 2.4 instead of the name of a standard, that is barely followed by anyone, at least in the consumer segment.

That is why I opted for the simple “Rec.709” and “sRGB” as the defaults. We can assume a shorthand for the displays that have been in common usage for a long time, and not introduce unfamiliar terminology to consumers. For example, most of those that have been using P3 (at least in a projection context, that is) will assume it is a gamma 2.6 unless otherwise specified. If it’s “HDR P3”, then someone would say P3 PQ or something like that. But we don’t usually include Gamma 2.6 otherwise because it’s the most common standard for projection P3.

Those people that don’t know what BT.1886 is also won’t know the subtleties between IEC 61966-2-1:1999 and Gamma 2.2. Hence why I don’t currently even mention the transfer curve for the “basic” use cases. Professionals, however, need to know exactly what transfer curve is being used. Specificity matters.

In this case, I for sure prefer BT1886 :grinning:
Rec709 ODT makes people think that it has something to do with Rec709 curve.
Because there are sRGB ODT and Rec709 ODT. And since their primaries are the same, it’s logical to (mistakenly) assume that their names tell us about their curves.

I think, Gamut + display EOTF are the most clear and even educational names for users.

1 Like

The notes from meeting #146 are now available.

1 Like

Would using the term “piece-wise” for sRGB help? Something like:

Main Spec Name - EOTF Value

  • DCI-P3 (Projector) - 2.6 EOTF
  • Rec. 709 / Rec. 1886 - 2.4 EOTF
  • Display P3 (P3-D65) - 2.2 EOTF
  • sRGB (Display) - 2.2 EOTF
  • sRGB Peace-wise - ~2.2 EOTF (IEC 61966-2-1:1999)

Perhaps not as it might be too different from how it’s listed in other areas, but just a thought. I think “EOTF” is a better term than “Gamma” but I understand that might be too technical for some users.

I think “1886” can be used as long as it’s adjacent to the familiar “Rec. 709” everyone knows and “2.4” is listed somewhere. Certainly not a bad thing to promote the term.

EOTF, a function, can’t be just a number. If I get it right, Gamma in these cases is not a loose term, but the name of a Greek alphabet letter Y, that is used as a power function.
So EOTF, a function, is a Power function. If not Gamma, it still has to be something.

I may be wrong, but I suspect using “~2.2” as a name for sRGB played a major role in confusing sRGB EOTF and power law gamma 2.2.
One can reasonably think, that sRGB curve is a power of not exactly 2.2, but a power of ~2.2.

While I don’t believe there is a chance to have “piece-wise” in the name, I would personally use this in naming. This math term is not known (at least in my native language) for the most artists.

So it would be a sort of a “I’m sRGB, but don’t select me because you’ve read all computer monitors are sRGB, I’m also this word from math, google it at first”.
Like a precaution thing no to choose it blindly for every sRGB monitor.

Best would be to not ship the piece-wise encoding function as ODT.

8 Likes

As much as I don’t like the idea, now is a good time to take a decision if any!

1 Like

Personally it matters not to me what they are called, just so that the ones used are represented… and the names are unique and well defined somewhere.
BT1886 is different from gamma 2.4 in the low end.

I’ve been under the impression (per LightIllusion’s guide) that encoding for BT.1886 was still expecting a pure 2.4 power EOTF, and optional differences are how a particular display may be configured to display that signal with respect to black level, white level etc. which has varying opinions on whether that should be implemented depending on who you ask / who’s doing the calibration. Hope I haven’t been misguided on this but happy to hear otherwise.

1 Like

from Portrait Displays:

This article is talking about display calibration / adjustment, not the original signal encoding which I believe is still a pure 2.4. Plus on self emissive (OLED) displays BT.1886 ends up resulting in pure 2.4 anyway.

1 Like

Yes the black offset in ITU.BT 1886 is an optional way for displays to implement the brightness and contrast knob.
It is flawed because the operations are the wrong way around.l, I believe.
I think from a practical point of view one can say that gamma 2.4 with a black offset of 0.0 is a valid choice of encoding images to the spec.

1 Like

They deliberately made it that way round, and then later changed the black handling in BT.2100 HLG to work the same way. So somebody somewhere thinks that is the way it should work. I seem to remember Tim Borer from the BBC was an advocate of this approach.