[argyllcms] Re: Capture One Profiles
- From: Gerhard Fuernkranz <nospam456@xxxxxx>
- To: argyllcms@xxxxxxxxxxxxx
- Date: Sun, 07 Jul 2013 19:17:34 +0200
Am 07.07.2013 18:18, schrieb Ben Goren:
On Jul 7, 2013, at 9:02 AM, Alex Jamison <a539jamison@xxxxxxxxxxx> wrote:
If I 'flood the zone' with colors in the areas of special interest that the
resulting profile will give more accurate results in those regions?
Hopefully.
More sampling will permit the profiling software to build a better map of that
part of color space.
However, if the camera suffers from metameric failures in that part of color
space, no amount of sampling will get you accurate results.
The point is that the reflectance spectra of the (additional) patches should
still be representative for the object which you eventually intend to capture.
Patches which provide only a metameric match to the real object colors are
sub-optimal, if not counter-productive in some cases.
And one should avoid contradictory patches. If one patch sais that a particular
RGB triple should be mapped to a particular XYZ triple, and another patch sais
that the same RGB should be mapped to different XYZ numbers, which one should
be honored by the profile then?
If you're building your own chart, you first want as many different spectra as
you can get, and then you want as many different patches covering as large a
gamut as you can get.
For printer profiling you can't have enough patches, but I'm not convinced that more and
more arbitrary patches are generally better for input profiling, since some fill-in
patches which are not representative for real object colors might be contradictory to
representative patches then, and non-representative patches may furthermore lead to a
biased profile (particularly in case of a rather "stiff" model with a rather
low number of parameters).
Best Regards,
Gerhard
Other related posts:
