[argyllcms] Re: Using a RGB-Camera as colorimeter
- From: Ben Goren <ben@xxxxxxxxxxxxxxxx>
- To: argyllcms@xxxxxxxxxxxxx
- Date: Fri, 3 Nov 2017 11:48:30 -0700
On Nov 3, 2017, at 11:17 AM, Charlie (Redacted sender "chpine2" for DMARC)
The people who do Raw Therapee make .dcp profiles from ColorChecker shots
submitted to them per their instructions. I gave them a shot for the Fuji
X-Pro2, and they would not use it until I did it again properly. Would one of
these serve as a profile for the purpose under discussion here?
I very, very much doubt it.
Even so-called "best practices" for photographing charts is woefully far from
adequate...and even the best theoretical 24-patch target is a minuscule
sampling of color space.
Reflective charts certainly have a vital place in a colorimetric reprographic
workflow, but that place isn't making camera profiles. Note, for example, that
different lenses have different spectral absorption profiles, and even a
pinhole "lens" still causes mirror-box flare. Optimally, you'd be projecting a
spectrum directly onto your camera's sensor to get your base profile -- and
then separately getting the spectral characteristics of your lenses, your
illuminant, etc., and using all of those to generate your profile.
At its heart, the job of a profile is to translate the spectral response of the
camera's sensor to that of the human eye's spectral response. No matter the
technology used (ICC / DNG / matrix / LUT / whatever), the math ultimately
winds up being the logical equivalent of such a function.
Spend some time looking at a plot of a Standard Observer function (any of them
will work for this exercise), a similar plot of a camera's spectral response,
and the reflective spectra of the ColorChecker patches. You should begin to
appreciate how those mere 24 samples really aren't enough for the task.
...not to mention, of course, that there are colors that humans can see that
cameras can't and vice-versa; there're undefined and undefinable parts of any
such function...which means that there's going to be compromises, even in the
ideal case...and navigating those compromises is where a lot of the art comes
(To be fair, the mismatches are at the edges of the gamut, the super-saturated
colors not typically found in reflective media. But, even then...do you bend
the gamut to a best-fit match, or clip it, or use an hybrid approach? Different
tasks call for different techniques.)
Other related posts: