Research Of Image Segmentation Methods Based On Fuzzy Systems Theory

Image segmentation based on fuzzy technique has been widely and deeply studied, gradually becomes a branch of image processing field, and obtains a series of profound and effective result. However, there are some deficiencies and problems that needed to be improved in most researches. This paper deeply investigates image segmentation based on fuzzy system theory from three aspects: emphasizing strictness in using mathematics devices; improving the existing approach and algorithm; and introducing new fuzzy techniques in image segmentation domain.Emphasizing strictness in using mathematics devicesThresholding segmentation technique based on measure of fuzziness is undoubtedly one of the most popular researches in image processing, and is also one of the most intensive appliications of fuzzy techniques in image processing. In various relative researches, how to choose an appropriate formula calculating the degree of fuzziness and how to define new measure of fuzziness, to enhance the efficiency of image segmentation are dealt with frequently. People had always considered the problem of selecting dynamically threshold as a problem about measure of fuzziness of fuzzy sets, but they did not notice that it is essentially comparison of elements in an algebra structure, and thereby regarded mistakenly order homorphism (partial ordered set to totally ordered set) as order isomorphism. This results in randomness of selecting degree of fuzziness or mathematics devices. Beginning from fuzzy system theory and algebra theory, this paper makes a detailed theoretic analysis for nonconsitency of degree of fuzziness with the point of view of an order structure, and proposes a uniform criterion of selecting formula of degree of fuzziness, to avoid randomness of using mathematics devices.More recently, thresholding segmentation technique based on fuzzy partition entropy had been widely concerned. However, the same as above, the most researchers focus on how to define fuzzy partition entropy, or how to obtain the optimal...