clipka <ano### [at] anonymousorg> wrote:

> If by "analyzing" you mean anything that is related to human perception
> (as opposed to anything that is related to physics), then a somewhat
> perceptually uniform colour space would certainly be advantageous, such
> as CIE L*a*b or CIE L*u*v, or the even newer DIN99 series of colour spaces.

Probably true.
I recall spending much time observing your experiments with color gradients and
gamma.
(I think it would be nice to "remap" a color space from HSV to expand the very
narrow red region, and fix any other perceptually non-uniform characteristics.
Just for fun  ;) )

As for my original intent, my "analysis" is an effort to rapidly get a more
pleasing result from sampled colors for use in color maps when trying to
procedurally texture things.  You may recall my lamentations about trying to
replicate the beautiful honey-yellow coloration of a wood floor, since my color
maps only produced pale, washed-out results that weren't anywhere near what I
wanted.

A pipe-dream feature would be to have the RGB values of an area selected in the
render window (like the current re-render selected area feature) written to an
output file, which could then be used as a prototypical color mapping data set.

The edges of my brain's tentacles suggest that 3D fourier transform of that data
may well highlight the major contributing "3D" RGB vectors better than any
visualized histogram in a color space would.
Or I may just be overcomplicating the task.  ;)

I'm also wondering what algorithms various automatic pallette-generating apps
use (Android / Windows Store apps for interior decorating, selecting paint
colors, etc.)

Post a reply to this message