Re: Haxachrome gamut shape
Re: Haxachrome gamut shape
- Subject: Re: Haxachrome gamut shape
- From: Graeme Gill <email@hidden>
- Date: Mon, 14 Mar 2005 12:50:58 +1100
- Organization: Argyll CMS
Roger Breton wrote:
Please excuse my ignorance but I did not suspecct that there could be a
difference between the two? Can the same be said for CMYK "raw" vs
"reachable" device gamut?
I think so.
> Well, I'd really like to see one.
I have a VRML of it, but it's rather large (about 2Meg).
See <http://web.aanet.com.au/gwg/test8.wrl>.
Note that this is a plot of the gamut surfaces of each colorant
combination, where the surface is artificially colored to indicate
where surfaces come from.
Why the difference is beyond me? Are Hexachrome gamuts fundamentally
different from CMYK or CcMmYk gamuts? Isn't it "just" a matter of inverting
the A2B table?
The basic problem is this:
And N channel device has a raw color response that can be described
as a 2^N vertex, N dimensional hypercube. Attached to each vertex is the actual
color that colorant combination (ie. 0 or 100% of each colorant)
produces. Colorspace is 3 dimensional, so the effect is that of "folding"
the N dimensional hypercube into 3 dimensions. Topologically this cannot
be done without "crossing" edges, sides and volumes etc. over one another.
What this means in effect is that you generally have interpenetrating "edges"
and "volumes" of the N-dimensional hypercube forming the points of the gamut
extremes. (My specially colored version of the Roland 8 color gamut surface
is designed to show this aspect.)
The points where these different edges and volumes intersect are points where
there are device value "aliases" for the same color value. These aliases can and
will have quite different device values. What this means in a typical color reproduction
system (one that uses a sparsely defined table such as a B2A to translate between
color values and device values, and interpolates the in-between values), is that
you can't just "leap" between one device alias and another, which is something
you'd ideally like to do to be able to follow the outside of the device gamut.
To be practical, you have to find a path that has device value continuity,
where linearly interpolating the device values gives you color values that
are what you intend at the input color. Such a usable path will probably
have to depart from the gamut surface, and this means that it is almost
certainly not possible to achieve the theoretical maximum gamut, and
have useably accurate color. Almost certainly there are trade-offs between
color accuracy and gamut, and gamut in a particular color area and gamut in
another area.
Some N color profiling approaches use a device computational approach
(ie. generating some device values by "taking" value from an input CMY
of CMYK device values), and this approach (exactly analogous to the
K = min(CMY) type of black generation approach) doesn't give you an
optimal gamut volume. For instance, some of the gamut surface defining
points for the Roland 8 color are those involving combinations that
are rather non-intuitive, eg. light ink and green, or light ink and orange,
or partial combinations of two inks that in the device computational
approach are seen as "exclusive" combinations, and not ever generated.
I hope some of the above is comprehensible :-)
Graeme Gill.
_______________________________________________
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