GENETIC-BASED FUZZY HIT-OR-MISS TEXTURE SPECTRUM FOR TEXTURE ANALYSIS

Abstract

A new method using a fuzzy hit-or-miss transform recognition procedure, which measures degrees of fit of specified patterns within an image, is proposed for texture analysis. For a given texture, three optimal texture patterns (structuring elements) dynamically generated by genetic algorithms are used to inspect the degrees of fit by using the fuzzy hit-or-miss transform, respectively. The distribution of these degrees of fit, converted into a texture spectrum, is used as the texture feature for texture analysis.