Ben Goren
2014-10-08 17:05:17 UTC
Graeme,
Is there any chance that input profiling could be expanded to support both spectral reflective measurements of the reference charts and .sp illuminant readings of the source used to photograph the chart?
The current workflow is ideal for copy work / art reproduction. Any differences between the illuminant used and D50 get effectively "baked into" the profile. No matter the character of the illuminant, so long as you always use that same illuminant with the same profile, the resulting Lab values will (within tolerances) match the Lab values you'd generate from a contact spectrometer reading.
But, as ideal as that is for copy work, it's a serious problem for general-purpose camera profiles. Unless your illuminant just happens to be a perfect match for D50, you're once again "baking into" the profile the differences between your illuminant and D50. Further, the general aim in general-purpose photography isn't to emulate a spectrometer, but rather a spectroradiograph -- you're photographing the light as much as you're photographing the objects. If the reference XYZ values for the chart are generated not against D50 but instead against the actual illuminant used to photograph the chart, the result should (again within tolerances) be a similar (quasi-)"absolute" rendering of the scene.
Typical workflows will still involve the development of the camera's data being responsible for handling white balance conversion, which presents all sorts of problems -- but the type of "absolute" profile I described would still make such a workflow "less worng." It would also open the possibility of always developing with the same white balance as used for profile creation, and then using colorimetric processing, presumably using a .sp file, to handle white balance. It would also, I think, in theory, permit the guesstimation of the illuminant's spectral characteristics from a picture of a suitable chart -- but that's perhaps better left for another day.
It is, of course, perfectly possible today to use Argyll's standard tools to calculate XYZ reference values for a chart given a suitable .ti3 and .sp pair, and I'm already experimenting with exactly that. What I'm hoping is that it wouldn't be much effort from you to make that process transparent to the user.
I'd offer to write the code, myself...but I still haven't tried to dive into it....
Thanks,
b&
Is there any chance that input profiling could be expanded to support both spectral reflective measurements of the reference charts and .sp illuminant readings of the source used to photograph the chart?
The current workflow is ideal for copy work / art reproduction. Any differences between the illuminant used and D50 get effectively "baked into" the profile. No matter the character of the illuminant, so long as you always use that same illuminant with the same profile, the resulting Lab values will (within tolerances) match the Lab values you'd generate from a contact spectrometer reading.
But, as ideal as that is for copy work, it's a serious problem for general-purpose camera profiles. Unless your illuminant just happens to be a perfect match for D50, you're once again "baking into" the profile the differences between your illuminant and D50. Further, the general aim in general-purpose photography isn't to emulate a spectrometer, but rather a spectroradiograph -- you're photographing the light as much as you're photographing the objects. If the reference XYZ values for the chart are generated not against D50 but instead against the actual illuminant used to photograph the chart, the result should (again within tolerances) be a similar (quasi-)"absolute" rendering of the scene.
Typical workflows will still involve the development of the camera's data being responsible for handling white balance conversion, which presents all sorts of problems -- but the type of "absolute" profile I described would still make such a workflow "less worng." It would also open the possibility of always developing with the same white balance as used for profile creation, and then using colorimetric processing, presumably using a .sp file, to handle white balance. It would also, I think, in theory, permit the guesstimation of the illuminant's spectral characteristics from a picture of a suitable chart -- but that's perhaps better left for another day.
It is, of course, perfectly possible today to use Argyll's standard tools to calculate XYZ reference values for a chart given a suitable .ti3 and .sp pair, and I'm already experimenting with exactly that. What I'm hoping is that it wouldn't be much effort from you to make that process transparent to the user.
I'd offer to write the code, myself...but I still haven't tried to dive into it....
Thanks,
b&