We address the issue of low-level segmentation of color images. The proposed approach is based on the formulation of the problem as a generalized Voronoi partition of the image domain. In this context, a segmentation is determined by the definition of a distance between points of the image and the selection of a set of sites. The distance is defined by considering the low-level attributes of the image and, particularly, the color information. We divide the segmentation task in three successive sub-tasks, treated in the framework of Voronoi partitions: pre-segmentation, hierarchical representation and contour extraction.