Hi Erik,
I think Torger answered this question with a definite yes, with the usual cautionary qualifications. I think I can see why: DNG profiles are designed to allow us to take any input and turn it into any output via double roundtrips to HVS with stopovers for additional PP (go wild with those sliders, Ken), although compared to a nice, pure linear matrix it seems at first glance to be a little bit like cheating. This is where I wish we had a bit more clarity from the cognoscenti:
When the goal is trying to objectively characterize HARDWARE quantitatively, as opposed to subjectively achieving pleasing output qualitatively - answering questions that depend on CFA Spectral Sensitivity Functions like sensor gamut, color discrimination, separation, orthogonality - shouldn't we be able to calculate that information straight from the forward 'compromise' matrix? Hasn't anyone come up with some practical such metrics yet, besides useless SMI? When someone says 'that camera has poor yellow discrimination', shouldn't we be able to point to a ratio or other combination of terms in the compromise matrix and say yay or nay? Please show the math in your answer.
Jack
The question of why the metrics are so bad is a good one.
One thing I always noticed when visiting Gretag to discuss some new measurement device or do a meeting on camera profiling was the lack of any images on the wall, and the very bad lighting.
What people who work in repro usually do is create their own targets out of a bunch of random materials; the vendor targets have many colors but actually very few pigments, which is why any images shot of them with different cameras can be "superposed" as soon as the pigment recordings are aligned for a given illuminant. So if a programmer says that the profiles can be matched in the sense of matching CC24 to CC24' under incandescent lights, that can be true even if the two cameras will then give completely different renditions in practice, or even on a target made with random materials.
I don't think there is any really hard maths involved (disclaimer, I used to be a mathematician so maths-shaming may not work with me). But there is a fundamental incomprehension of what the hardware should be doing -translating percepts into recordings- and also there is a refusal to keep track of the computational precision of the devices and conversion pipeline. The whole color management model was designed for the scanned-image print world, and it will take another 20 years to catch up with digital capture.
BTW, in my experience 90% of color issues go away immediately even in mixed lighting if you establish a profile in the light of your shot, which is the real advantage of the small colorchecker targets. In the light of the shot means where the subject is, not where you are. I would suggest that people who have color issues start out by doing this rather than junking a whole camera system.
There are some papers out there about what camera sensors do, but in fact - at least when I stopped looking a few years ago - there aren't that many. Nowadays the industry is AFAIK basically riding on the coat-tails of the phone sensors, and that is where all the industry research money is getting spent, although I assume that car sensors are now going to be an even larger market with every car having 10 sensors, some of which will need very high dynamic range.
Camera color is not a research priority. The last major conference I went to, in the non-geriatric crowd (under 30) there were chinese-chinese, taiwan-chinese, american-born-chinese, chinese grad students from the US, chinese from Canada, chinese from the UK, chinese-chinese grad students from the UK, some indians from the US, some japanese, all from both sexes, a bunch of male and female europeans and a lot of aging US geeks, but not one single american white male under 30. In a phone-camera industry meeting I went to a year or so ago there were few women, but also not a single young white male from the US. As we all know, white males are over-represented in the US computer industry, so their complete absence is indicative of the absence of money and career prospects. On the other hand, color engineering clearly is one tech industry which does not have a diversity problem.
Edmund