Notice of Meeting - ACES Gamut Mapping VWG - Meeting #17 - 6/18/2020

Meeting Summary:

  • @nick walked the group through the BLGs he posted on ACES central for use in Nuke to help evaluate under different output transforms: Display Transform BLGs for Nuke for Gamut Compressor assessment
  • @Alexander_Forsythe suggested adding in just a 2.2 gamma transform- so you’re looking at the scene referred imagery - take the display transform out of the equation
    • Similar to what Thomas’ notebook originally did
  • @matthias.scharfenber - would it help to also have a very ‘simple’ display transform too - simple S curve and gamma etc (CAM16 mapping to display space?)
  • @LarsB says ITU recommended gamut mappers are based luminance take no account of saturation (hard clip to BT.709 boundary)
    • BT. 2446 has the three different methods
    • BT. 2390 - different rolloffs, bezier curve, hard to control - tonemaps on the PQ code values
  • When evaluating the images in a display-referred manner, a few points were raised:
    • @Thomas_Mansencal would like to avoid the few people deciding approach
    • @carolalynn points out that this isn’t necessarily about aesthetics, but about how “correct” the image is
    • @joseph points out that same viewing environment is really important, controlling that is almost impossible if you can’t have people in the same space
  • @LarsB pointed out that it’s not about the look, it’s about functionality: can you grade this image better, can you comp this image without issues.
    • Can we come up with a set of scenarios for compositing & grading, so we can have a comparative test? Farm for easy examples on ACES central and in our group, figure out if we can use our current set of images, and what we’re lacking
  • @Alexander_Forsythe reiterated that we should look at the numbers and analyze as much as possible first before we visually evaluate
  • @joseph has some tests he’s shot, will try and find good frames to share
  • Could also shoot some more headlights for chromatic aberration etc
  • @matthias.scharfenber described a test scenario with incandescent light on one side, with overlapping LED on the other - @joseph will try to look into shooting it
  • Where do we put the algorithm in the pipeline turned into a long discussion:
    • Concern for implementation and maintenance if we made it part of the ap0->ap1 transform for both ACES contributors and DCC developers
    • We will start a list of pros and cons on DropBox Paper to be added to by the working group