Hello, I often find that a lot of time in my raw editing is spent trying to tame wild wild out of gammut colors, and that a lot of the tools in Rawtherapee that can do so either have side effects or seem to occur after tone curves in the pipeline, even the gamma referred rgb mixer, which creates artifacts of various kinds, such as color inversions and flattened out or pre-clipped regions. Also sometimes the negative values created from out of gammut dark colors tend to create issues. What I would like to know how to do is find the numeric values in the color matrix inside a dcp file so I can left multiply it with a desaturating matrix, and use that new color matrix whilst still in linear color space for a better starting point, even if it means that the other tools inside the DCP arenāt applicable anymore, such as the LUT.
You can use DcamProf to read and write to dcp files; or you could convert the files to DNG and read those (they are just a TIFF, see the DNG Spec to index into the relevant tags). Though you have to know a bit about the make up of the profile not to get into trouble.
For instance Adobe in recent years has been using desaturated matrices to get into XYZ (with no negative terms in DNG converter and their own dcps) - then spreads things out again via the LUTs. DcamProf does something similar with default settings.
OK, now I have used your link to see what is inside dcp profiles, and have found four matrices, colormatrix1, colormatrix2, forwardmatrix1, and forwardmatrix2. I donāt know which does what, and what Iād want to perform the desaturation matrix multiplication operator to get my desired expanded gamut.
Update: So Iāve looked up the info in the DNG spec you linked, and it was clear that the colormatrix1 and colormatrix2 were just the different sides of the dual illuminant. The forward matrix part of the text was beyond my understanding, but Iāve read elsewhere that it seemed to not matter much, so safe to ignore?
Keep reading Matrices (and related LUTs) are for the two spread-out illuminants indicated in the file, usually A and D65. They are interpolated based on the estimated correlated color temperature at the time of capture, in mired units.
If you are looking at an Adobe dcp, ColorMatrices are mainly used to determine CCT and take you from XYZ to ācameraā space as-is. ForwardMatrices are the ones you are interested in, take you from white balanced data to XYZD50 and are designed to potentially also work with LUTs, see the previous proviso. Good luck.
Also of note is the companion dcamprof page, Making a camera profile with DCamProf. In the Appendix thereās a writeup on DCPs that describes the workflow they support.
Thanks for pointing this out! Iāve referred to the dcamprof documentation countless times but I never noticed the detailed description of the DCP pipeline!
I hear you. I got so tired of trying to ārecoverā highlights and saturated color that I ended up making a custom DCP profile that transforms my cameraās native color into Arriās Log-C Wide Gamut color space. Hereās an Arri Alexa still (from the web) to give you an idea what that looks like:
The nice thing about doing it this way is that Arriās log tone curve and wide gamut color space are documented standards meaning that itās simple to create or appropriate 3DLUT color transforms to shape the gamut and tones. The cool thing about doing this in the raw profile is that your exposure, white balance, etc, are all still happening in linear color, but the image is mapped to a film like log curve for editing giving much higher precision in the shadows for editing.
You can also use any DNG file if you like the colors and use it as a DCP profile on your imageā¦just load it instead of a dcp fileā¦just as a quick way to see what it might possibly look likeā¦
Great idea. Fitting dynamic range into range of image without tonal distortion side-effects was another issue I was having. Any links to tutorials? I got my cameras and I got a x-rite color checker passport.
Also, it would be great for better mixing Rawtherapee raw processed with Prores, as I can just use log profile conversions to get to ArriLog in Resolve.
Interesting⦠how did you define the tone curve for lumariver/dcamprof?
I downloaded the JPEG, it definitely re-energizes the S-curve. You got me thinking to do something similar with ACES2065-1, although that is still scene-referredā¦