A colour texture image segmentation strategy which integrates region and boundary information has been described. The algorithm uses the contours of the image in order to initialise, in unsupervised way, a set of active regions. Therefore, colour texture of regions is modelled by the conjunction of non-parametric techniques of kernel density estimation and classical texture features. Afterwards, region compete for the pixels optimising an energy function which takes both region and boundary information into account.
The method has been quantitatively evaluated on a set of mosaic images. Furthermore, results over real images riches in colour and texture are shown and analised. The results demonstrate the effectiveness of the proposed algorithm in estimating regions and their boundaries with high accuracy.