CIE
Division 8 - TC8-03: Survey of Gamut Mapping Papers
Montag & Fairchild (1997)
Four mapping techniques applicable either to lightness or to chroma are
presented in this paper. The approach adopted here is to evaluate the effects
of the proposed techniques on lightness and chroma separately, whereby
CIELAB hue angle is kept constant. Figure 1 shows the principles of the
following four proposed techniques.
Scaling & clipping (SC)
As can be seen, this technique is a combination of linear scaling and clipping,
whereby the individual implementations lie between the two extremes of
clipping all out-of-gamut colours onto the boundary and of compressing
all colours so as to fit into the reproductionís gamut. Several combinations
of scaling and clipping were used and are denoted by the distance between
the two extremes.
Knee function (KF)
This technique, as well as the following two, is a piecewise linear mapping
technique. It consists of two linear mappings, the first of which has a
slope of one and the second compresses the remaining range to the maximum
of either the image or the device gamut (this is similar to method (b)
by Herzog and Müller (1997)).
Gentile et al technique (Gea)
This method was proposed by (Gentile et al., 1990) and also consists of
two linear mappings: the first has a slope between one and that of the
complete linear compression and the second slope is determined by the first
lines intersection with the cut-off line determined by the function x =
max (output).
3 Segments (3S)
As the name suggests, three linear mappings are used here, the first segment
being the same as in the KF technique, the last being a clipping and the
middle segment connects the other two (this is similar to the piecewise
linear method proposed by Sara (1984)).
Figure 1 Mapping techniques evaluated by Montag and Fairchild (1997).
Paired comparison was used to evaluate the various implementations
of the above techniques. This was done by gamut mapping simple rendered
images to artificial gamuts simulated on the CRT on which the original
image was displayed as well.
The evaluation suggested that the best results were obtained when image
gamuts rather than the medium gamut were used as the original gamut. For
chroma mapping on its own, the best reproductions were made with the SC
algorithm, whereby all out-of gamut colours were clipped onto the boundary.
Lightness mapping was tested for two cases: one where the top end of the
lightness range was limited (here the 1/3Gea technique and the maintaining
of saturation worked best) and another where the bottom end was limited
(here the maintaining of chroma and the clipping of all out-of-gamut colours
was preferred). The authors have also noted that the use of a colour appearance
model space would be beneficial, as it could prevent hue shifts caused
by the mapping when CIELAB was used. Further it was suggested that combined
chroma and lightness mappings should be investigated.
Last updated: 17 August 1999 by Jan
Morovic