Re:Kodak CMM in Photoshop
Re:Kodak CMM in Photoshop
- Subject: Re:Kodak CMM in Photoshop
- From: Chris Murphy <email@hidden>
- Date: Mon, 23 Apr 2001 23:15:08 -0600
Chris Cox writes:
>
ProfileMaker has also has problems with some of it's own data
>
scaling. They were supposedly working on a fix (but I don't have
>
3.1.4 to test).
I have yet to encounter the problem with 3.1.4 profiles and the Adobe
(ACE) engine; but it still occurs with other CMM's.
>
And with some profiles, you CANNOT get a "correct" result due to the
>
way the tables are built (ie: if white has to be interpolated, it's
>
almost certain to be wrong....).
*This* is what I have a major gripe with. I don't care if the profile
doesn't contain correct data regarding white or not. White should NEVER
be interpolated. The PCS should be irrelevant. White point adaptation
should be irrelevant. 255r,255g,255b should always, always, always,
always convert to non-image/non-ink values (255,255,255 for RGB and
0,0,0,0 for CMYK). CMM's should know this and do an end run around the
profile -i.e. IGNORE the profiles themselves whenever a pixel with a
value of 255,255,255 or 0,0,0,0 is encountered while using the relative
colorimetric, perceptual or saturation rendering intents. Any kind of
effort on interpolating white should only occur with absolute
colorimetric.
At this point there are so many screwed up profiles floating around, even
embedded in images and archived, that even if you fix the problem in ICC
profiles themselves it is NOT ENOUGH. The CMM needs to take control of
this situation. It's the fastest way to do it, it ensures the same
results regardless of who made the profile, and it's the only way to deal
with legacy profiles and prevent this cockamamie problem from happening.
>
Adobe is working with everyone (ICC, OS makers, and vendors) to at
>
least identify the problems, but it's a slow process -- and some
>
vendors seem to have trouble with the basic math.......
Yet another reason to not put the solution to this particular problem
into the hands of some vendors. Deal with it in the CMM. I would imagine
the "conversation" a CMM would be having during this process:
1. Source profile is RGB, mmmkay.
2. Image is RGB, mmmkay.
3. Rendering intent is perceptual, mmmkay.
4. Destination is CMYK, mmmkay.
5. First pixel value is 255,255,255; immediate convert to 0,0,0,0 - do
not consult source profile. Do not consult destination profile. No
further computation on this pixel.
6. Second pixel value is.....
If the rendering intent is absolute colorimetric, then yes I would expect
the source and destination profiles to be consulted and for any necessary
interpolation/adaptation to occur. But NOT with the other rendering
intents.
Chris Murphy
Color Remedies (tm)
Boulder, CO
303-415-9932