CIE
Division 8 - TC8-03: Survey of Gamut Mapping Papers
Ebner & Fairchild (1997)
The study described in this paper is of particular importance, as it results
in experimental data on which gamut clipping can be based. It consisted
of a psychophysical experiment designed to "find the perceptually smallest
distance from a point in color space to a closed surface of color (gamut
boundary)". This was done by having as the originals three kinds of images
(abstract [a uniformly coloured square with or without a border], data
[a uniformly coloured chart] and figurative [a uniformly coloured cartoon
car]) which were coloured with 25 colours from the surface of the original
CRT gamut. Observers were then asked to "adjust the color of the image
on the left-hand side of the screen to make it look as close as possible
to the image on the right-hand side of the screen", whereby the original
image was on the right and the left-hand image could only have colours
from the simulated reproduction gamut. The reproduction gamut was the gamut
obtained by intersecting the gamut of a colour laser printer with that
of the CRT. Even though this meant that the gamut shapes of the two media
were different, it seems from the figures in this paper that the lightness
ranges were made to be equal. As there was no significant difference between
the results for the different image types, they were all pooled together
for further analysis.
For a given out-of-gamut colour, the experimental results were then
modelled by a weighted combination of the following three vectors: (a)
constant L* vector, (b) centroid vector (towards L*=50 on the lightness
axis) and (c) minimum distance to gamut surface vector at the original
colourís hue angle. The experimental results were also compared with the
work of Katoh and Ito (1996) and good agreement was found with the coefficients
set to (1:2.4:0.6). This result agrees with the findings of Katoh and Ito
in as much as it gives least weight to chroma but disagrees with it by
claiming that hue needs to be preserved more than lightness. In spite of
some differences, these results indicate some maturity in the understanding
of gamut clipping and either of the above coefficient sets seem to be good
choices for situations where it is sufficient to overcome gamut differences
by clipping.
Last updated: 17 August 1999 by Jan
Morovic