Graeme Gill wrote: > Photo papers use three colorants (CMY), so the resulting spectra are > always a combination of those three spectra. The real world has many > different spectra that appear to be the same color to our eyes, and > such a chart is only exercising one choice of them. > > So as a result it is (theoretically) possible to model a typical RGB > sensors response to such a chart exactly, even if the sensors > spectral sensitivities aren't much like that of the human eye. > > Exposed to real world spectra, the profile for such a sensor created > this way might be rather inaccurate. On the one hand (when profiling a scanner for scanning photos or film) it is an advantage that the spectra of photographic media have only 3 degrees of freedom, since this eventually enables a bijective mapping from XYZ to a single corresponding reflectance/transmittance spectrum on the medium (for a given illuminant, and for in-gamut colors only of course), which in turn usually also enables a unique, bijective (though non-linear) mapping between scanner RGB and XYZ. But on the other hand the reflectance spectra of photographic media seem to be pretty metameric wrt. real-world objects, thus a camera profile generated from an IT8 target (or any other target chemically printed on photo paper) may result in a color cast when applied to shots of real-world objects. Eventually it also depends on the used light source and the camera, of course. Visually judged, the resulting observer metamerism (camera vs. human vision viewing target vs. real-world objects) seems to be not a too big problem in conjunction with daylight (outdoors) or a tungsten light source, but it seems to be rather problematic e.g. in conjunction with a fluorescent light source. Regards, Gerhard