Recap:
- @Pablo_Garcia - used the DCTL implementation on a current project with good results, tweaked per clip for some more creative uses
- Known bugs:
- Shadow rolloff still has effect at 0
- Hexagonal mode inverse is incorrect right now, implementation is possible but not pretty
- Proposal to move forward without the hexagonal mode, no real objections
-
@daniele presented an issue with the reds - start from a “perfect” camera that fills the spectral locus, compress. Quite a noticeable red gap
- You don’t see it with the RRT, but visible still in T-cam
- Also found an artifact issue on tanh with inverting all negative values
- Objective testing:
- @KevinJW : how important is P3 based on Daniele’s tests around saturated reds? Could start with P3 and invert to see where it lands as a test gamut to compress.
- @daniele : round trip inversion tests - how many can we do before we get unacceptable errors (what is that metric?)
- @joseph - time dependent, willing to test around color purity in arri measurements
- What happens if you use the RICD & Harald’s LED spectra measurements? Similar to Daniele’s already demonstrated tests
- @nick : for testing should we reset / set all values to the same number and let the user find what works best for them?
- @Alexander_Forsythe : should we do a pair comparison test or other objective user test measurements?
- Defaults need definition - what should the mapping do in a default state - these are our next steps!