The importance of the above coding is not only in being a more appropriate way of storing the L*u*v* data, but also that the whole gamut compression algorithm is based on it. The method proposed by the authors is to find the colour in the original deviceís gamut which is furthest out of the reproduction gamut and then scaling the original gamutís unit tetrahedra so that this colour gets onto the boundary of the reproduction gamut. The authors have recognised the naiveté of this approach and have also suggested that the image gamut should be used rather than the device gamut. Non-uniform compression is also considered, however it is suggested that this would "no longer preserve similar perceptual relationships among the colours in the input image".
The method described above will inevitably result in excessive compression,
as the gamut is being compressed uniformly and linearly. All but one region
of the colour space will be compressed too much and this will be most noticeable
if the shapes of the original and reproduction gamuts are dissimilar. The
argument of loss of perceptual relationship when gamut compression is carried
out in a non-linear manner would be valid if the gamut mapping colour space
would be uniform. However, in CIELAB or CIELUV colour differences around
the neutral axis (for example) are more perceptible than those between
highly chromatic colours in which case a non-linear mapping could better
preserve perceptual relationships.