[argyllcms] Re: Device link gamut mapping: how to do it right?

Nikolay Pokhilchenko wrote:

As I understood, the -G"image.gam" mapping in collink is equivalent to simple 
mode (-s
-ip -op), except additional compression changes based on gamut overlaps. 

Well, no. Simple mode uses the gamut mapping encoded into the source profiles 
A2B
table, and the destination profiles B2A table. It is whatever the profile 
creation
software puts in there. The -g and -G modes in contrast, create the gamut 
mapping
specified by the collink parameters.


So, it's some
equivalent to standard perceptual mapping. 

Well, no. The gamut mapping is whatever is specified with the -i option. It 
could
be perceptual, or it could be something else.


You say, that there is the "the neutral axis
matching (which includes the white point)" so, it's contain a kind of relative 
linking.


It is a relative linking if a relative based gamut mapping is selected with 
(-g/G) -i,
and it's not if a non-relative based gamut mapping is selected.


It seems to me, that there is relative scaling before compression from input 
profile to
output in collink -G.


I'm not sure what you mean by relative scaling. If a relative colorimetric based
gamut mapping is selected, then the color is all made relative to the white
points of the respective devices, and the white points are mapped from one to
the other. For a gamut mapping intent like perceptual, neutral axis mapping
is performed (to match the dynamic range of the devices), and then 
hue/saturation
mapping is applied.


Is there relative scaling of "image.gam" in collink before
compression too? 

The gamut should be created in a way that is compatible with the
intent being used (kind of messy to use I know, but that's how it
is at the moment), so if a relative to white intent is being used,
then a relative to white gamut should be created.


I can't belive, that there is -ia intent necessary instead of -ir in
tiffgamut preceding collink -G. I've experimented a bit, and noticed, that while
tiffgamut -ir there is less clipped areas in output images, than wile tiffgamut 
-ia
before collink -G linking.


I would expect there to be some subtle differences, because the gamut is 
different.
But for use with collink perceptual intent, a relative colorimetric gamut in
the same color space (ie. CIECAM02 Jab space with the same viewing conditions)
is the correct combination.

Graeme Gill.

Other related posts: