CIE Division 8 - TC8-03: Survey of Gamut Mapping Papers

Hoshino & Berns (1993)

An overview of different gamut mapping techniques (all of which are described in the preceding sections) is given in the beginning of this paper after which the following four lightness mapping techniques are evaluated using a paired comparison method:

linear lightness mapping (compression to 95 per cent (minimum necessary to fit into reproduction gamut) and 90 per cent of the original scale were used)
soft clipping (a technique similar to the one used by Stone and Wallace (1991) was used, whereby the first 75 per cent or 95 per cent of the reproduction lightness scale from white were left unaltered and the remaining 25 per cent or 5 per cent respectively were linearly mapped onto the darkest black of the reproductionís gamut (Figure 1). Again clipping to 95 per cent and 90 per cent of the original scale were used)

Colourfulness mapping was not evaluated in this experiment and chroma clipping with constant hue and colourfulness was used. Note, that the gamut mapping space was based on Huntís colour appearance model (Hunt, 1991).

The results from the paired comparison experiment in which printed reproductions and original images on transparencies were compared, suggested that the successfulness of the compression depends on the amount of compression. Where only small amounts of compression were needed, 95 per cent soft clipping resulted in the best reproduction, whereas for large amounts of compression the results were image dependent. For large amounts of compression, the softclipping approach was better for images of high to medium lightness, but was found to be worse for dark images as it introduced an artefact referred to as fluorence. Fluorence was defined as the instance "when the brilliance of the stimulus exceeds that of its surrounding or comparison stimuli" (Evans, 1974), i.e. when it appears to be fluorescent.

Figure 1 Lightness mapping techniques.

In terms of interpreting the results of this paper, it would have been useful to know the colourfulness ranges of the test images used and it is of use to keep in mind that the differences between original and reproduction gamuts were relatively small (compared to gamut differences in the experiments of this thesis). Nonetheless, the evaluation of lightness mapping techniques described here provides useful data for setting up a gamut mapping algorithm and should be taken into account.

Last updated: 17 August 1999 by Jan Morovic