Closed KevinJW closed 1 year ago
As reported by @nick-shaw
When turning on/off highlight desat we see a break in Hellwig 2022, independent of the choice of tonescale curve.
The choice of curve does shift the location the break occurs at
Gamut compression and focus bypass also trigger differences
After some investigation we noted that surround illuminant adaptation appears to be one cause of the colour bias as would be expected when the surround does not match the input white point.
The plot above is using the "live from params" LMS matrix, which exaggerates the effect. Turning off gamut compression (which should be doing nothing with a neutral ramp) removes the break, but shows how wonky the curves really are.
What are the curves here? The red one not being monotonic is quite something!
Linear ramp -> ACEScct to ACES. v23 from @alexfry repo. Settings as in the screenshot. I would like to know if others see the same or it’s my system.
Yes, @nick-shaw, I see the same thing with "Live from params", but I never use it typically. I don't see that craziness with the red channel with "Stock" params.
I think I've fixed this in v024
Looks like the issue was related to the variable "discount_illuminant" vs "discountIlluminant"
The discontinuity occurs when HK mode was on 'in the middle' solution is don't do that.
As reported by Pekka when feeding a pure grey RAMP onto both models we see a slight colour difference.
Thoughts are to look for a matrix that doesn't quite row sum to 1, or an extra CAM, etc.
Kevin