Fingerprint segmentation based on fuzzy theory

This paper deals with a new algorithm for fingerprint segmentation based on fuzzy theory. Combined a new feature, GR (gray range) with other traditional pixel feature, three standard patterns are constructed and classification is realized by closeness degree and MMP (maximal membership principle). A selection step is added to remove much of unnecessary region. Instead of removing the whole remaining region, we focus on a balance between necessary region and unnecessary one by the joint entropy of the two images. Even for fingerprint with worse quality, the proposed algorithm can obtain a satisfactory segmentation by adjusting parameter or the size of window. Furthermore, all parameters are adaptable and the method is applicable into any fingerprint database. Extensive experiments demonstrate the efficiency of our algorithm.