I wanted to try and process the raw NEF files that my camera gave me using my own code, so I wrote some. Then I put it into a library and now you're reading the readme for that library.
I want to try and do color stuff myself, but if I get stuck I will probably look at palette (github)
- Everything on brucelindbloom.com. Bruce is a color scientist.
- 4.2.1 of this pdf
- "Completely Painless Programmer's Guide to XYZ, RGB, ICC, xyY, and TRCs"
- poynton Color FAQ
- https://www.strollswithmydog.com/determining-forward-color-matrix/ and the rest of the stuff there.
- white balance "tint" https://www.dpreview.com/forums/thread/4385471
If you know this, think you figured it out, or otherwise just want to email me. Please. My head is exploding and I must know the answers. gen@nyble.dev.
- What is "unity" as referenced in poynton's Color FAQ here?
Things I do not currently understand but hope to in time
-
Why, when we chromatically adapt during the XYZ -> sRGB transform, does the image come out looking like darktable's "as shot" whitebalance? It's CCT is ~5972K and my xyz->srgb transform should be D65 white, so. It's weird that I'm getting the as shot considering I should be moving the white point? right???? And if this is correct, how do I get darktables result? What are we doing differently?
-
It seems that my XYZ -> sRGB and chromatic adaption is off by 4.8% on the blue channel. why? how do I fix this? Is this all rounding related? Is 5% more blue enough blue to make ~6000K look like D65?
The tests and benchmarks require raw files that I don't want to stuff into this repository because
of their size. They are located here: https://nyble.dev/rawproc/testfiles.zip. Extract that to
tests. It should look like tests/raw/<lots of raw images>.
The three major types we recognize are u8, u16, and f32.
Whitebalance:
- BayerRgb: u8, u16, f32
- LinRgb