Problem with KisAbstractColorspace::difference

[Cyrille Berger]

> It uses deltaE which is an international standard, based (to a degree) on
> perceptual input from actual persons.
>
> However, it is most accurate when colors are close, so the extremes you
> describe are not handled very well.
>
> If we can find a way to provide better long distance accuracy while
> retaining close accuracy too - it would be great.

And I found a page where they seem to prove that deltaE is a bad measure 
anyway (I don't know if they are right or no, I don't know enought about 
color ;) ) :
http://www.aim-dtp.net/aim/evaluation/cie_lab/index.htm

And lcms use an outdated DeltaE function :
http://www.brucelindbloom.com/index.html?Math.html
New version might be more accurate, but they ask more calculation and I am not 
sure they answer correctly to our problem.

And for RGB, I think (abs(r1-r2) + abs(g1-g2) + abs(b1-b2))/3 might work, but 
equivalent function for CMYK will be wrong.

-- 
--- Cyrille Berger ---