Image segmentation using fuzzy correlation

A concept of correlation between two properties (fuzzy representations) of an image is introduced. A set of algorithms for image segmentation (both fuzzy and nonfuzzy) has been formulated. The spatial information is taken care of by the following measures: transitional correlation and within-class correlation. A relation between the correlation coefficient and the index of fuzziness is theoretically established and experimentally verified. The effectiveness of the algorithms is illustrated on images having different types of histograms.