[Openicc] [Fwd: icm profiles in debian]

[Kai-Uwe Behrmann]

Am 11.02.2013 19:25, schrieb James Cloos:
>>>>>> "KB" == Kai-Uwe Behrmann <ku.b at gmx.de> writes:
>
> KB> Can you please switch on the light for us all and elaborate, why
> KB> you think algorithmically profile generation is essentially right?
>
> The weight of continuing evidence has convinced me that stuffing a
> mathematically defined space into a file format designed for sampled
> data fails miserably.

ICC profiles can contain parametric curve elements, which are not 
sampled data. However this is a V4 feature and does not work with V2 CMMs.

> Linear interpolation; differing white points, adaptations, et al all
> add up to confusion, incompatibilities, unnecessarily reduced precision
> et cetera.

That reads like a problem statement. And we have the discussion here in 
parts because of that. Keeping profiles selectable in order to allow for 
various concepts and ideas people might come with, appears to me worth 
the effort put in fixing the ICC profiles.

> Use the definition of the spaces to convert between them and XYZ, and
> save the icc profiles for conversions between XYZ and sampled targets
> such as displays and printers.

Colour spaces are defined by spread values or formulas. But conversions 
between colour spaces are defined by the algorithms used in profilers 
and baked as a result into ICC profiles.

> And preferably in floating point, where possible.

CGATS like used by Argyll and other profilers should allow for that 
unsampled data. Btw. ICC profiles can contain conversion tables in 
floating point precission.

kind regards
Kai-Uwe