Different reproduction devices can have different sets of reproducible colours. These sets are called gamuts. The process of transforming colours from one device (or image) gamut to another is called gamut mapping. Gamut mapping has many technical issues to be considered: the used colour space, direction and magnitude of the mapping and whether and to which extent ingamut colours should be altered. Spatially invariant algorithms treat all the pixels independently on their position in the image. Spatially variant (local) algorithms allows a better rendition but introduces the problem of artefacts and/or haloing in the resulting image. In this paper we propose a spatially variant gamut mapping algorithm that creates virtually no artefacts nor haloing in the resulting image. We start from an analysis of the Retinex algorithm and devise proper functionals to build an algorithm which tries to maintain spatial ratios in the image while mapping it into the gamut and, at the same time, avoids all drawbacks of Retinex approaches. We suggest to perform the mapping in an RGB colour space rather than one of the perceptually more homogeneous ones. Although less homogeneous, we experimentally show that RGB colour spaces actually have better hue constancy according to a certain criterion.
Many image clustering algorithms use distance metric in the process of taking decision. When dealing with color images, a distance metric will be used to decide whether two pixels or regions are closed. Colorimetric distances proposed by CIE(Commission Internationale de l'Eclairage) are often used in Lab color space because it is a uniform chromaticity space. However, RGB color space is useful to image processing and instead of converting color image from RGB to another color space before processing, it might be interesting to have the same or better results without changing the color space. In our work, we implement different distance metrics and compare the result of k-mean clustering algorithm in RGB color space to the one in L*a*b* with the colorimetric distance. Two evaluation criteria have been used and we conclude that being in RGB color space and choosing adequately the distance metric, we obtain better segmentation results.