Three-Dimensional Control of Color Management Systems - Control of Color Imaging Systems: Analysis and Design - page 397

Clusters can be formed by running the vector quantization algorithm (Section
2.4.7) for the L*a*b* nodes or can be formed in or around the vicinity of the corners,
edges, and faces. In the entire color space, with this method, it is not impractical to
imagine more than 50 clusters being processed intelligently by the computer with 90
different gamut-mapping algorithms.
7.6.6 B LACK P OINT C OMPENSATION
RGB images often contain colors that are darker than the darkest color a printer can
make. Minimum color error can be obtained by mapping the out-of-gamut colors to
the darkest color the printer can make. However, in that case, all details in the
shadow regions may be lost. They often contain information the viewer knows
should be there (the folds of a dark coat, the strands of dark hair) and their absence
can be very disturbing. Therefore, an L* mapping technique for black point com-
pensation similar to the one shown in Figure 7.60 is often used for RGB images.
It retains the details for dark colors, though at a reduced contrast, and at the cost of
lightening some of the darkest colors (see Figure 7.63). For implementing the
black point compensation (i.e., input L* to output L* map along the neutral axis)
FIGURE 7.63 (See color insert following page 428.) Example of black point compensa-
tion (top left: original, top right: without BPC, bottom: with BPC).