Hello Roger,
thanks for your explanation, which I unfortunately cannot completely agree
with, as far as my understanding of the ICC-profile goes.
My understanding of the creation of an ICC-profile is as follows:
The target scan is analysed by "scanin" and the RGB-values measured for each
patch are related with the XYZ-values in the reference table. This is shown
in the associated Argyll-file ti3. Then a mathematical model is created by
"colprof", that relates the RGB-values with the XYZ-values, either with a
LUT or via a matrix, what have you. If a profile is assigned to a scanned
raw image, this table is used to transform the measure RGB-values into the
device independent XYZ- or Lab-values. Thus I do not understand your
statement, that " The profile calculation will be different so that the
results ARE the same, that is, RGB = 137, 140, 112 becomes Lab = 50,0,0."
This would mean that the transformation would not be unique. I would assume
that a different RGB-set will correspond to a different Lab- or XYZ-value
according to the mathematical model. If the target is correct, a different
set of RGB-values must also correspond to a different set of Lab-values --
all of course within measurement errors.
Next step is the transformation into the Adobe RGB (1998) colour space. For
this, the above calculated (device independent!) XYZ- or Lab-values are
transformed according to the prescription that defines the Adobe RGB colour
space, i.e. how XYZ transforms into an RGB-values. In the image header of
the new output image it is now stated that the RGB-values in the data file
are to be interpreted as Adobe RGB, which again is device independent. If
the image is displayed on a monitor, the ICC-profile of the monitor is used
to transform the Adobe-RGB-data in the file into device-dependant values,
which are sent to the electronics, in order to produce an image on the
screen, that is colour managed and looks the same on all (calibrated)
monitors.
I hope, this is at least roughly correct. Otherwise my understanding of
colour management would be completely wrong :-(.
Best wishes
Hermann-Josef
--- Begin Message ---Du can nur shrift eine bisset worten auf Deutche (only write a few words in
- From: "Roger Breton" <graxx@xxxxxxxxxxxx>
- To: <argyllcms@xxxxxxxxxxxxx>
- Date: Tue, 14 Mar 2017 14:47:08 +0100
German), I am not fluent in deutch, aber Ich liebe experiment :-)
Here is the main concept of an ICC profile, regardless of Silverfast or
Photoshop or Argyll: die profile arbeit ist to construct a mathematical
relationship between the device values und die device-independent (Lab)
values.
So, suppose patch GS10 is expected to be L = 50, a = 0 und b = 0.
Das ist der Referenz value. (You can look this up into the IT8 referenz
file) Once the profile "execute", after been created, no matter what the
starting RGB values are, (to some extent) the result must be Lab = 50,0,0.
So, if the scanner RGB values are R=120, G=100 und B = 135, once the profile
is created, in Argyll or in any other profiler, this combination of RGB
values is expected to be transformed to 50,0,0 in Lab.
It is possible that the profile nicht give you that value, 50,0,0, because
of many reasons, but the results should not be too far, maybe 49,1,2, you
see? (Sehen sie?)
The profile becomes a machine to transfer RGB values to Lab in a fixed way.
But what if the scanner RGB values are R=137, G=140 und B=112?
The profile calculation (the mathematical coefficients in the equation that
transforms RGB to Lab behind the scene) will be different so that die
results ARE the same, that is, RGB = 137, 140, 112 becomes Lab = 50,0,0.
You see? That is why if you "overexpose" or "underexpose" the war IT8 scan,
does not change the expected results, the IT8 Lab values, as long as you
always scan a little overexposed or a little underexposed. You scan in a
FIXED way.
Now, I suggest, for the best results, study Silverfast configuration so that
you don't "apply" any ICC profile -- turn ICC Color Management Off.
Until you understand how Silverfast ICC Color Management arbeit.
Und experiment creating the scanner profile mit Argyll und Assign die
profile when you open the image in Photoshop.
The visual results should be sehr gut!
MfG / Roger
-----Original Message-----
From: argyllcms-bounce@xxxxxxxxxxxxx [mailto:argyllcms-bounce@xxxxxxxxxxxxx]
On Behalf Of Hermann-Josef Röser
Sent: Monday, March 13, 2017 10:38 AM
To: argyllcms@xxxxxxxxxxxxx
Subject: [argyllcms] AW: Re: AW: Re: AW: Re: ICC-profile shift histograms
Hallo Roger,
soll ich auf Deutsch schreiben??
regardless of how you adjust the raw scan, over-expose, under-expose,profile is supposed to give you the same results.
the
I think I have seen that the results differ, but I will check again.
I would not understand this, because in the reference file also L is
specified. So if the target scan is darker, there must be an adjustment to
compensate for that to get the desired L. Am I misunderstanding something
here? It will give the same and and b (i.e. same colours), but not the same
brightness. Is that true?
Is there a way to show images here? If not, I will post some explanatory
examples on my ftp server for retrieval.
Many thanks for your efforts and best wishes
Hermann-Josef
--- End Message ---
