I've been reading this great article on how digital cameras convert raw data to sRGB images or to XYZ space. From the article (specifically section 7) and the DNG specification I understand that for Adobe DNG files
Forward Matrix (F) * Raw Channel Multipliers (D) = CAT(Scene White Point --> D50) * Inverse Color Matrix (C^-1).
So applying either side of the equation to the camera RGB values of the illuminant for which the forward and color matrices are optimized should result in D50 XYZ. I wanted to test that for my Fuji X-T4. I use both forward and camera matrices 1 for illuminant A and matrices 2 for illuminant D65. I assume that the CAT is linear Bradford (as recommended in the DNG specification). I obtain D as 1 / (Color Matrix * Illuminant XYZ) -- because applying the Color Matrix to the XYZ value of the illuminant gives the camera RGB values of the illuminant, it's inverse gives the multiplication factors required to map the camera RGB values to (1, 1, 1).
Although both sides map the RGB values of the illuminant to D50 XYZ, the matrices do not match. That means that RGB values other than those of the illuminant will be mapped to different XYZ coordinates.
Then I took the camera profiles in "...ProgramData/Adobe/CameraRaw/CameraProfiles/Adobe Standard/" and tested the equation above for all Fuji cameras which had profiles in that folder. The equation does seem to hold for cameras made before autumn 2013 but not for most cameras after that date.
So I'm curious about what has changed and why the equation doesn't hold now. I'm not aware of any additional factors that have to be accounted for. Is a different CAT applied? Which one?
(Forward and Color matrices 1 and 2 can be obtained from the metadata of a .dng file or from the camera profiles (.dcp) in "...ProgramData/Adobe/CameraRaw/CameraProfiles/Adobe Standard/".)
Thanks for your help!
2
Replies
AdChoices