Lars Borg <bo...@...>
Well, I had planned to write a longer mail, but it got sent prematurely.
Here is a longer version:
Is there any interest in creating WCG/HDR-ready ICC profiles within OCIO?
The current set of ICC profiles from OCIO (I looked at the ones created
1.0.3) has some short-comings in current workflows and that we should
consider addressing. Here's my list:
1. The profile’s PCS space is too small, just Lab.
The ICC specification states that compliant implementations (CMMs)
shall constrain (clip) conversions to the ICC Lab range. The ICC Lab range
is 0..100 for L, and -128..+128 for a and b. This space was sufficient for
It is not sufficient for say Rec. 2020. For Rec. 2020 the green corner
is outside ICC’s Lab range. This is easily shown in a ColorSync plot.
Other spaces that support colors (already in SDR mode) wider than ICC Lab
includes ACES, ACEScc,ACEScg, ARRI LogC, Sony S-Gamut,
Now it might not matter to you as today all your colors are inside the
P3 space and (SDR) P3 fits completely within ICC Lab range.
Another aspect that saves the day is that not all CMMs are ICC
compliant. Some CMMs support intermediate values outside the Lab range.
But now you’re relying heavily on deviations from the spec.
2. The profile is an SDR profile, not supporting any HDR content or HDR
Lab maxes out at ACES diffuse white (100). When applying RRT ODT to
ACES HDR shots, this (in a compliant CMM) clips the specular highlights.
Constrained XYZ gives us 2x headroom. Not much but should give us less
However, we can use ICC profiles in non-constrained mode, extrapolating
into HDR space. Marti Maria (lcms) suggested this years ago, and we’ve
been doing this since 2006. All but 3D LUTs can be extrapolated. We’re now
routinely creating HDR profiles for HDR TV, log cinema spaces, etc.