colour-science
Implement relevant changes regarding "ALEXA Log C" IDTs.
@JGoldstone has updated ALEXA Log C IDTs and we need to look at implementing the relevant changes on our side.
References
- https://github.com/ampas/aces-dev/pull/98
I did a quick rewrite of the code from @JGoldstone's IDT generator, and compared it with what Colour produces:
ACES IDT LogC 0.9, EI 500 = 19.4102494417
Colour LogC 0.9, EI 500 = 19.4102861128
ACES IDT LogC 0.9, EI 800 = 21.6936400245
Colour LogC 0.9, EI 800 = 21.6934858954
ACES IDT LogC 0.9, EI 1000 = 22.9156705374
Colour LogC 0.9, EI 1000 = 22.9155157995
ACES IDT LogC 0.9, EI 1600 = 26.1717608061
Colour LogC 0.9, EI 1600 = 25.8352040602
At EI 1600 the difference is quite significant. The Colour function seems to top out at 1600, returning an error at higher EI. The Goldstone function is unconstrained.
Excellent @Nick-Shaw!
It makes an even stronger case to implement this new generator.
I would be keen to have an official statement from @JGoldstone saying that ampas/aces-dev#98 is the canonical official reference.
We can assume for now that it is but this typically goes back to section II - D in the ACES - Retrospective and Enhancements paper.
Nick, the canonical document is always the internal logC.R file, but I would say that the canonical external reference is the OCIO-config code that went in after 1.0.3, which was then echoed into aces-dev I believe. There had already been a 1.0.3 release of aces-dev, without an in-sync release of OCIO-config 1.0.3, alas, so I guess the 'official' aces-dev release that contains it will be whatever is after their 1.0.3 release.
Enclosed is a writeup of the change that I sent to the developer of logC.R before making the change that became part of OCIO-config 1.0.3.
shorterLogCLinearizationChangeJustification.pdf
