Re: How do one generate profiling patches (The hen or the egg)
Re: How do one generate profiling patches (The hen or the egg)
- Subject: Re: How do one generate profiling patches (The hen or the egg)
- From: Roger Breton <email@hidden>
- Date: Wed, 23 Apr 2003 08:46:02 -0400
Thomas wrote:
>
OK, maybe I didn't explain my point clearly enough:
>
>
Suppose I want to generate my own patches for whatever purpose (like Bill
>
Atkinsons TC 9.18 target for ProfileMaker). I'd build the CMYK (or RGB)
>
patch file with defined color values. Presumably I'd have to do this in CMYK
>
(or RGB for RGB profiles).
There isn't anything liminting you to building your custom targets in CMYK
only. If you want to experiment quickly, start by modifying an existing
target. Say, TC2.88. What you do is, in a text editor, modify the RGB values
defined in the Reference file "TC2.88 RGB Ref.txt". Use whathever patch
definitions you like. Then, use ColorLab to generate a TIFF file of that
target. Next, print that target. And measure it with Measure Tool. Next
comes building the profile with ProfileMakerPro. You see, you use PMP's
algorithms to generate a profile from you custom patch definition.
>
How do I generate a reference file for this new target? I need to be able to
>
build a profile for it...
See above.
>
If my understanding of profiling is correct (which it may or may not be) the
>
actual readings of a printed target are compared with a reference file. Thus
>
a gamut for the device is established, and at the same time the resulting
>
profile know what Lab coordinates the defined CMYK values will land on, and
>
have a basis for calculating the values falling between the defined patches.
I think you got that right. Except the first part. To me, the actual
readings are not "compared" with the "reference" values contained in the
reference file. What all CMS packages do, to various extent, AFAIK,
depending on how far reaching their color science and mathematical models
are, is to build statistical relations between device values and Lab or XYZ
values (actual measurements). To the extent that the statistical relation
found proves to be strong between the two sets of variables (as revealed by
correlation index or other indices) then we know we have an accurate way to
go from device to Lab, A2B. There are all kinds of techniques to extract
this relation from the data.
To get the inverse, from Lab to device, B2A, is less trivial. But it is
along the same line. People talk of "inversing" the table.
>
What I don't get is the "double digital" target (CMYK and Lab that must have
>
a correlation) where you print one and compare the readings from this with
>
the other (lab reference).
Not sure I understand your question here. There are no two targets: only
one.
>
It's the link between the two (CMYK or RGB patch file, and Lab reference
>
file) I don't get...
Sheer mathematical modeling, Thomas. If you want to read up furhter on this,
I suggest you pick-up copies of Gaurav Sharma's Digital Color Imaging,
Color Engineering edited by Phil Green, Color Technology for Electronic
Imaging Devices by Henry R. Kang. There are more and more good monographs
coming out on the subject of color management. (Abhay Sharma?) the problems
with these books is that, some only gives you an overview of an area of
color because they are written as a collection of "texts" for an
undergraduate or graduate course on color management. Presumably, the
instructor builds on the the text in class, allowing interaction with the
subject through a class setup -- essentially picking up on the teacher's
brain. Usually, what happens with these books, like many books in general,
is that the coverage tends be limited to the area of interest defined by the
author. If the author only devotes three pages to printer modeling in a 356
page book, that's it. To satisfy my personal appetite for learning on this
stuff, I've had to buy many books that I keep revisiting as my knowledge of
color expands. In all events, if you're interested of learning more about
color modeling, as I think you are, you may have to build your own color
puzzle from bits and pieces gathered from published litterature and
participation in newsgroups like this one. There is also sci.eng.color which
is color-related and worth visiting regularely.
Roger Breton
Laval, Canada
email@hidden
_______________________________________________
colorsync-users mailing list | email@hidden
Help/Unsubscribe/Archives:
http://www.lists.apple.com/mailman/listinfo/colorsync-users
Do not post admin requests to the list. They will be ignored.