Colour reproduction for reflective images

Abstract

Because of differences between the colour gamuts of different printing devices, often some colours of an image are not producible by the output device of choice. To handle this problem, gamut mapping is used to make certain that every input colour has a corresponding printable output colour. Gamut mapping is normally done in a trichromatic device-independent space, such as the CIEXYZ, CIELAB or CIELUB colour spaces. The colour coordinates in such spaces can only be computed by assuming a source of illumination. Unfortunately, satisfactory results obtained for one viewing illuminant do not guarantee good results for other illuminants. Using colours specified as spectral reflectances can solve some parts of this illuminant dependency problem.

The basic concepts of colour image reproduction in reflectance space are developed in this thesis. The major reproduction steps for reflective images, including device characterization, gamut mapping, and backward colour transformation are examined. For device characterization, a colour lookup table (CLUT) with nonuniform sampling should be used for better accuracy. Reflectance space gamut mapping based on objective measurement and subjective measurement were investigated. A novel algorithm that preserves the fundamental component of reflectance is developed. Compared to the usual mapping algorithm, projective mapping, the new algorithm consistently produces results with small colour differences between reproduction and original colours for multiple illuminants.

The essential concepts for developing gamut mapping that based on appearance matching in reflectance space are developed. Two gamut mapping mapping algorithms based on perceived colour attributes are provided. Finally, to reduce the high cost of cell extraction associated with nonuniformly sampled CLUTs, a novel algorithm for cell-finding is developed. This algorithm improves the performance of the backward transformation, which must be done once per image pixel.