Am 14.03.2017 um 21:18 schrieb Mike Horvath:

> Do you have an idea what those images are supposed to show? Is it the
> spectral locus? I've only plotted the sRGB gamut so far (which is shaped
> like a skewed cube), not the spectral locus.

Before I can answer that question, we may have to first agree on a
definition of "spectral locus".

According to my understanding of the Wikipedia article on "spectral
color", the "spectral locus" would be the locus of all monochromatic
colours, i.e. colours comprised of only a single wavelength.

In a 2D chromaticity space (a "colour" space that does not care about
absolute brightness) such as CIE xy, that would be the famous "horseshoe".

In a 3D colour space, it would be an extrusion of that horseshoe, traced
on an arbitrary locus of equal brightness in that colour space, extruded
along paths of constant chromaticity, up to the locus of zero brightness
in one direction and up to the locus of infinite brightness, in the other.

For example, in CIE xyY colour space it would be a "cylinder-ish" shape
(having a cross-section identical with the familiar CIE horseshoe
shape), oriented along the L axis, starting at L=0 and extending to
infinity.

On the other hand, in an RGB colour space it would instead be a
"cone-ish" shape (with a cross-section also reminiscient of the CIE
horseshoe, albeit possibly distorted depending on the angle at which you
cut), encompassing the positive legs of all colour axes, with its apex
at R=G=B=0 and extending to infinity.


According to that definition, the images are clearly /not/ supposed to
show the spectral locus.


Instead, from what the other guy is writing, it is my understanding that
the images are /supposed/ to show the locus (or rather, selected points
from that locus' boundary) of all possible /pigment colours/, under a
poorly defined illuminant (from the description my guesses would be E,
the equal-energy illuminant), in CIE L*a*b colour space whith a poorly
defined whitepoint (my guess would be D65 or D50).

The side view looks reasonably convincing (I don't recall ever having
plotted this shape in CIE L*a*b space). Note that towards L=0 the locus
appears to converge to (*a,*b)=(0,0), which seems to agree with the
mathematical definition of the colour space.

The top view, on the other hand, has some features that make me
suspicious; but they might just be artefacts resulting from a different
scaling along the L axis than used for the side view, or I might be
seeing ghosts.

Post a reply to this message