[en] When a colour image from one colour reproduction medium (e.g. nature, a monitor) needs to be reproduced on another (e.g. on a monitor or in print) and these media have different colour ranges (gamuts), it is necessary to have a method for mapping between them. If such a gamut mapping algorithm can be used under a wide range of conditions, it can also be incorporated in an automated colour reproduction system and considered to be in some sense universal. In terms of preliminary work, a colour reproduction system was implemented, for which a new printer characterisation model (including grey-scale correction) was developed. Methods were also developed for calculating gamut boundary descriptors and for calculating gamut boundaries along given lines from them. The gamut mapping solution proposed in this thesis is a gamut compression algorithm developed with the aim of being accurate and universally applicable. It was arrived at by way of an evolutionary gamut mapping development strategy for the purposes of which five test images were reproduced between a CRT and printed media obtained using an inkjet printer. Initially, a number of previously published algorithms were chosen and psychophysically evaluated whereby an important characteristic of this evaluation was that it also considered the performance of algorithms for individual colour regions within the test images used. New algorithms were then developed on their basis, subsequently evaluated and this process was repeated once more. In this series of experiments the new GCUSP algorithm, which consists of a chroma-dependent lightness compression followed by a compression towards the lightness of the reproduction cusp on the lightness axis, gave the most accurate and stable performance overall. The results of these experiments were also useful for improving the understanding of some gamut mapping factors - in particular gamut difference. In addition to looking at accuracy, the pleasantness of reproductions obtained using various algorithms was also looked at and a strong and positive correlation was found between these two properties. It was also shown that the reproductions made with GCUSP were pleasant in isolation. which makes it a very good candidate for a standard universal gamut mapping algorithm. (author)