Re: Rendering Intents
Re: Rendering Intents
- Subject: Re: Rendering Intents
- From: bruce fraser <email@hidden>
- Date: Sat, 31 Aug 2002 09:46:06 -0700
At 6:46 PM +1000 8/31/02, Graeme Gill wrote:
michael shaffer wrote:
This would beg the question as to "why?" That is, ... a conversion
between gamuts of similar size would ask for very little compression ... so
why does perceptual fail? Are there more fundamental differences 'tween the
2 intents?
I can think of two reasons that may explain this.
One is due to a fundamental limitations of the ICC profile
design, that the gamut mapping is implemented primarily in
the output table (B2A table), without any knowledge of what the source
gamut actually is. I would therefore presume that a perceptual
B2A table in most profiles, will compresses regardless, thereby
producing an inferior result when no compression is actually needed.
It's not really an ICC profile limitation, it's a corollary of the
"smart profile/dumb CMM" architecture of ICC color management. The
upside to the architecture is that we can obtain very similar results
from different CMMs, and hence we have interoperability and an open
standard. What we lose is the ability to do adaptive mappings based
on the real gamuts in use at a given moment.
A sensible gamut compression scheme does so with the full
knowledge of the source and destination gamuts, and (if it
worked well), one might expect a perceptual and colorimetric
intent to produce very similar results if the two gamuts
are similar.
To do that, I think you'd need a smarter CMM, and simpler profiles. A
profile only has knowledge of itself and the PCS. The gamut mappings
we build into profiles are based on an assumed source gamut (usually
undocumented). The packages that allowed you to choose a specific
source gamut on which to base perceptual rendering seem to have
fallen from favor -- none of the current major players let you do so.
The ICC spec also lets you implement perceptual gamut mapping in the
source profile. But the same problem applies: unless you know both
the source and destination gamuts, an optimal perceptual mapping
isn't possible -- you just have to make a middle-of-the-road
assumption.
To get a color management system that produced adaptive perceptual
mapping, wouldn't the intelligence to do so have to be in the CMM?
Profiles are just data...
It would be fascinating to see someone build such a system to see if
it actually produced better results. Meanwhile, I use Relative
Colorimetric with Black Point Compensation about 90% of the time.
Bruce
--
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.