[argyllcms] Re: Histograms shift when ICC profile is applied
- From: posts@xxxxxxxxxxxxxxxxxxxxxxxxx
- To: argyllcms@xxxxxxxxxxxxx
- Date: Tue, 12 Jul 2016 10:47:27 +0200
Ben, thank you for your reply.
This would principally be a matter for your scanner software and
especially the channel / white balance and / or target white point.
Exposure is also going to be a factor.
There was no white balance involved in the software and also the
exposure time should not be an issue since scan of the IT8-target and
scans of the slides were made with the same settings, i.e. the same
exposure time per pixel.
The simplest and easiest and quickest way for you to judge the
quality of your workflow would be to scan a ColorChecker and compare
the recorded values against those expected. If you're getting
single-digit (or lower) DE errors for all patches, call it a day. If
not...if not, the nature of the errors will point you in the
direction of what to focus on next.
This is exactly what I had done and brings me to the next question: I
had used profcheck to get information about the ICC-profile and got
"Profile check complete, errors (CIE94): max. = 9.958780, avg. =
0.678745, RMS = 1.149767", which is quite good as far as I can judge.
Then I applied this very profile to the scan of the IT8-target and
stored this image as Adobe1998. Then I used "color transformer 2" in
imageJ to convert the RGB to Lab values. For each patch I measured the
dE between the Lab-values thus obtained and the Lab-values in the
reference file coming with the target. The result was very different
from what profcheck reported: Maximum of dE was 25, minimum 1.8 with
an average of 14.9. I would have thought that this experiment should
reproduce the profcheck output but it does not. Why is that?
Other related posts: