RE: "Proper Gamut Mapping"
RE: "Proper Gamut Mapping"
- Subject: RE: "Proper Gamut Mapping"
- From: email@hidden
- Date: Fri, 14 Dec 2007 16:57:07 -0500
> Todd Shirley writes (Fri, 14 Dec 2007 12:44:30 ):
>Are you referring to the fact that a conversion
>using standard ICC profiles requires 2 profiles because
>each profile can only describe one device, and therefore the
>destination profile is converting from a "non-specific source
>gamut" because it, by itself, does not "know" the source? How
>is this a problem, considering you must use a source and
>destination profile for any conversion?
It seems to me that two issues are being mixed together here which, if
separated, may implicitly provide some clarity.
The first issue pertains to the word "conversion" which implies a change of
state. In a color management context it typically means going from a source
device space (i.e. RGB) and ending up at a destination device space (i.e.
CMYK). How one does this is an implementation question. It can be done via
two separate src/dst profiles connected via the Profile Connection Space
(PCS), or via a Device Link in one fell swoop. In the context in which I've
described it, it is more a mathematical operation that relates to color.
Notice that I haven't used the word "gamut mapping" so far.
The second issue is the question of color quality. Not all color
transforms are of equal quality; sometimes this difference arises from the
different "secret sauces" the various vendors employ. Other times it has
to do with First Principles. This latter point is demonstrated if we
consider the two typical means by which Device Links (DVLs) are built.
DVLs can be simply concatenated together from separate source & destination
profiles; or you can build a DVL from the ground up starting with knowledge
of both the source and destination devices. The latter approach offers the
promise of a far superior transform than the former; simply put, you're
starting out with more information. Version 2 ICC device profiles are
typically built without knowledge of the "other" profile taking part in the
paired conversion, hence a simple concatenation of V2 profiles will
probably produce mediocre color (in comparison to an ab initio
construction). The version 4 ICC spec improves on this situation by
defining the PCS better and providing a Reference Medium Gamut which is a
common target in the PCS for both src & dst devices. But the best of all
possible worlds (from a First Principles perspective) is to have the closed
system, i.e. knowing both the src & dst devices from the get go. Hence
DVLs built in this fashion offer the highest possible quality. Notice
again that the word "gamut mapping" didn't appear in this paragraph either.
There are a number of challenging technical aspects in building a color
transform, be it a discrete profile or DVL. It includes how well and
accurately the overall system is being modeled; the K strategy employed;
mechanics of constructing the tables, etc. But perhaps the most telling
stage is at gamut mapping (especially in RGB workflows) which is defined as
moving a color that is out of gamut to a place in colorspace where it is
in-gamut for a given device. How this is done is part of the "secret
sauce" for each vendor; some do it better than others. If you know both
the src & dst gamuts you can do a better job than when that pairing
information is not there. Indeed, morphing one gamut to another is called
"re-rendering".
> I get confused when you say "how can the source
> profile have any affect on the gamut mapping ?"
What Graeme says is true especially for V2 output device profiles: there is
nothing in the V2 spec which requires it to know anything about the source
device (again, V4 improves on this situation); nevertheless the B2A table
does gamut mapping even though in some sense it is "blind" to where the
data is actually coming from.
>..because the LAB numbers are limited to the source gamut
>the destination gamut is "informed" by the source gamut?
And what you say is also true - how the colors are placed into the PCS is
of course determined by the source device. And any B2A will ensure that
that color will be gamut mapped to somewhere on or in the gamut of the
output device no matter where its placed initially.
The point to keep in mind is that the more you know about the source &
destination devices at the time you're building a color transform (profile
or DVL), the better the job you can do.
Harold
TOSHIBA AMERICA BUSINESS SOLUTIONS
Harold Boll | email@hidden
" The sunlights differ, but there is only one darkness. "
Ursula K. Leguin
This message (and any attached files) is secret, confidential and proprietary and is intended solely for specific addressee(s). If the reader of this message is not the intended recipient, please notify the sender immediately by reply email and delete it and all backup copies immediately.
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Colorsync-users mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden