Good morning Roger,
thank you for your detailed instructions. Unfortunately I only have
PhotoShop Elements. So I cannot follow your recipe perfectly. However, I can
use other tools like RawTherapee or imageJ. I had analysed the delta E (I
have written an EXCEL macro to set up a batch for Argyll, which makes life
much easier for me than using the command line. And it always includes
profcheck to analyse the histogram of deltaEs and also does a second
iteration with the high deltas clipped.). That all looks normal.
The problem is the calculation of Lab-values from the target scan once the
profile is applied. Different tools (SilverFast, RawTherapee, imageJ, Bruce
Lindbloom's calc page) give different results.
What I have done is following the advice by Fraser et al. (2005), who
recommend editing the profile, if one is not happy with the results. One
always has to edit in the opposite direction as one would like the profile
to react. So in my case, if the profile produces results which are too dark
(bright) one has to make the target scan darker (brighter). To achieve that,
I use scans with different settings of the analog gain (which effectively is
the exposure time). Standard is 150, so if the profile produces shifts to
the left, I use an analog gain of e.g. 120, which reduces the shift. I hope,
this is the correct procedure.
I would have thought that there is one of the many switches in Argyll, with
which one could adjust the profile.
Best wishes
Hermann-Josef
--- Begin Message ---My advice,
- From: "Roger Breton" <graxx@xxxxxxxxxxxx>
- To: <argyllcms@xxxxxxxxxxxxx>
- Date: Sun, 12 Mar 2017 21:38:08 +0100
Try processing the raw image outside Silverfast to PROVE to yourself that
the ICC profile is doing its magic.
Meine frage ist that, how does the farben of der image appear to you in
Photoshop, regardless of the pixel conversion to follow?
Is it gut? Dark? Saturated?
It might be a fact of your scanner's response that certain farbe need
shifting the Red values down considerably to match the expected Lab values.
It's entirely possible.
Compare the performance of the scanner on the IT8 image itself, FIRST. What
is the DeltaE like?
If the pixels are converted by the profile, there has to be a gut reason.
This is easy to do. Just open up the raw IT8 scan (from which you create the
scanner profile) in Photoshop, assign the profile to the image, and compare,
one little patch at a time, the Lab of the IT8 image and the Lab of the
resulting image, after assigning the scanner profile.
(Make sure you select Absolute Colorimetric under Color Conversion Options
in Photoshop's ColorSettings).
It is not necessary to look at every patch of the IT8, just analyze those
patches that you find suspicious.
You have to find a way to build confidence in the arbeit of der profile.
Delta E color difference is very good for that analysis.
/ Roger
-----Original Message-----
From: argyllcms-bounce@xxxxxxxxxxxxx [mailto:argyllcms-bounce@xxxxxxxxxxxxx]
On Behalf Of Hermann-Josef Röser
Sent: Sunday, March 12, 2017 4:12 PM
To: argyllcms@xxxxxxxxxxxxx
Subject: [argyllcms] AW: Re: ICC-profile shift histograms
Good evening Roger,
Thank you for your answer.
I am using SilverFast HDR. I have the raw (linear) scanned image loaded,
which is gamma-corrected by HDR. Then the ICC-profile is applied and the
output image is converted to Adobe RGB (1998) colour space. In HDR I see
two histograms, the input histogram and the output histogram. If I turn off
colour management I see as input the gamma-corrected data without
ICC-profile. If I then switch on colour management with the ICC-profile of
my choice the colours in the displayed image change and the input histogram
changes in the way I described: it shifts left or right, in some cases it
stays put. Unfortunately one cannot show images here.
If I split the histogram into the three channels I can see e.g. that it is
the red channel, that saturates at dark values. The colours look okay.
I use imageJ to look at the pixel values. The output image in Adobe RGB
indeed has the red channel saturated at low values. So it is the pixel
values themselves that are changed by HDR.
Hermann-Josef
--- End Message ---
