On Automatic Selection of Fuzzy Homogeneity Measures for Path-based Image Segmentation.
Belén Prados-Suárez, Jesús Chamorro-Martínez, Daniel Sánchez · 2005
This paper is the continuation of a previous work were we proposed the use of t-norms to measure path homogeneity in path-based fuzzy image segmentation. Here, we introduce an approach to automatically select the homogeneity function, taking into account the characteristics of the region to be segmented and its contour. To this purpose we firstly approximate a characteristic of the region surrounding the seed, on the basis of the study of a set of fixed paths. Secondly, we study the relationship between this characteristic and the parameter that controls the behavior of homogeneity functions based on Weber's t-norm. Finally, we obtain a function that from the value of the computed slope, gives the value of the parameter whose associated homogeneity function is suitable to model the region.