I'm quite pleased with the conversions that Mathematica uses when importing (Canon) RAW image files, e.g., RI = Import["/IMG_6257.CR2", "RawImage"] and RJ = Import["/IMG_6257.CR2", "Image"].

(The former brings in the sensor data essentially exactly as my camera recorded it into the CR2 file; the latter converts that info into an attractive, “gamma corrected” image.)

I've been able to reverse engineer the conversions partially. E.g., I have formulas that accurately predict the Image values for each of R, G and B, given the RawImage data of nearby values. But other areas of greatest interest to me, specifically for points with low luminosity where the general formula is not computable, and the interpolation functions for areas of small details, escape me.

Can anybody steer me to documentation of what is going on? I've scoured the web for standard approaches, but it seems Mathematica is using some unique approaches to make nice, appropriate-contrast images.