The segmentation of images using spatially guided methods is directed by the relationships between pixels in an image region. These methods group regions that are homogeneous regarding a given image feature. The spatially guided methods are further subdivided into region-based, energy-based, and region and contour-based methods. The region-based methods include procedures such as the growing, the splitting, and the merging of regions. Some algorithms from this category include the J-segmentation algorithm, proposed by Deng and Manjunath,2 the gradient segmentation algorithm, proposed by Ugarriza et al.,3 and a multiresolution extension of the gradient segmentation algorithm, proposed by Vantaram et al.,4 among others. The energy-based methods attempt to minimize cost functions that model regions in the image. Those cost functions may be contour- or region-based functions. The regions covered by the functions evolve until a given energy model is minimized. Some algorithms in this category include the active contours (a.k.a. snakes), first proposed by Kass et al.5 and variants, such as the fast active contours algorithm proposed by Chan and Vese6 or the active contours without edges, proposed by Vantaram and Saber.7 Finally, the contour-based methods consist of different variants of the watershed algorithm. This algorithm considers a gray-scale image as a topographic relief, where the intensity of the pixels determines the corresponding height of that particular zone. The relief is then flooded in a simulation and the water flows to local minima and forms basins, corresponding to different regions in the image. Some examples of these methods include the work of Gao et al.,8 where watersheds are used to segment color images, the study of Hill et al.9 that uses a texture gradient to partition textured regions using watershed, and the method by Kim and Kim,10 where a multiresolution watershed segmentation using wavelets is presented.