An empirical comparison of statistical chromaticity models

The use of colour, as opposed to single band intensity data, has proved useful for tasks such as segmentation, tracking, and object recognition. Moreover, the use of statistical models for distributions of colour data corresponding to regions of interest in images is an effective way of exploiting the extra information provided by this higher dimensional information. In this paper, we analyse several approaches to modelling colour data and present results of an empirical comparison of models based on these approaches