Pattern recognition from compressed labelled trees of fuzzy regions

In this paper, a method of pattern recognition based on images split into a set of trees composed of fuzzy regions is presented. First, a fuzzy segmentation based on possibilistic c-means is carried out in the raster image. Fuzzy support have been defined from a first level cut. On each cluster, a fuzzy region is assumed to be a convex combination of sets with associated features. A set of sample trees is achieved from the application of the segmentation algorithm on characteristic objects. Then, a tree isomorphism to recognize is defined to recognize an object. At last, a new tree compression method is introduced to decrease the complexity when we have to manage with a large set of trees.