A Fuzzy Model for Skeletonization

An image can be viewed as an array of fuzzy singletons. Most of the existing thinning algorithms are performed on binary images to obtain a singe-pixel-wide skeletal representation, where an α-cut or thresholding on the fuzzy image is required. To avoid information loss and distortion generation, a fuzzy-logic approach directly applied on fuzzy images is proposed. First, a convolution on the fuzzy image by a bell-shaped function is carried out to obtain a nonflat smooth surface. Second, the ridge points are extracted by only one rule based on the topographic analysis of structure. Third, a membership function is used to generate values representing the degree of membership grades to the skeleton of objects. Finally, the major ridge points are linked together to construct the skeleton of the fuzzy image.