[argyllcms] Re: Version 0.53 - Gamut mapping warning message.

Roger Breton wrote:

My hypothesis is that there is a relationship between the "linearity" or
good behavior or mononicity of a device and it profilability, holding such
optical phenomenon as fluorescence constant (better yet, under no
fluorescence).


Well, certainly non-monotonic device behaviour is a sign of
a device that is difficult/impossible to profile. (This has
little to do with non-monotonicity of gamut mapping though !)
A more rational measure of the "linearity" of a devices characteristic
might be the "curvature" and "slope" of the device characteristic. Such
a measure might be extracted from the interpolated profile by using
the CLUT grid, or sampling the overall transform characteristic. Once
again though, there is the difficulty of establishing a meaningful
scale for the numbers that would be produced.

> When you cited gamut mapping and non-monotonic behavior, I 
was thinking that there could be a way to characterize a device degree of
predictableness out of a device measurements.


See above. Non-monotonicity in the gamut mapping is not related
to non-monotonicity in the device characteristic.

> You see, that's only a
thought, when generating a new *ti1 file for characterizing a device argyll
always come up with a device sampling scheme that accounts for the number of
patches specified on the command line. So, in essence, argyll is casting the
device into an abstract, idealized expected behavior. Why not report then on
the degree to which the actual measurements deviate from this idealized
behavior? 

That's not what happens by default. By default test points are spread
evenly in device space. If the -A parameter is provided, then yes, a
pre-existing profile can be used to estimate curvature, and to try and position
more test points where curvature is high. I haven't characterized the
benefits (or otherwise of this) though.

Graeme Gill.


Other related posts: