A Path-based Fuzzy Approach to Color Image Segmentation

Color is one of the most used features for image segmentation. However, two uncertainty problems arise in this scope: the color feature is imprecise by nature, and the boundaries between regions in real images are usually blurred. In this paper, we propose a fuzzy color-based image segmentation, where not only the obtained regions are fuzzy, but also the color feature used in this process. This way, our approach takes into account the blurriness of boundaries between regions, as well as the uncertainty associated to the color feature. On the one hand, colors are modeled through fuzzy sets related to the linguistic terms used by humans, providing membership degrees associated to them. On the other hand, we propose a path-based fuzzy segmentation technique, where the relationship between two pixels is quantified by means of fuzzy connectivity. This way, given a seed point, the connectivity between this point and the rest of pixels in the image can be used to obtain the corresponding fuzzy region.