Automatic bandwidth selection of fuzzy membership functions

Image thresholds can be chosen by minimizing the measures of fuzziness. However, there are no existing algorithms selecting the bandwidth of fuzzy membership functions automatically. In this paper, a new method for automatic bandwidth selection of fuzzy membership functions is presented. First, the peak locations are chosen from the histogram using the peak selection criterion. The suitable bandwidth is found iteratively between two peaks. Multilevel thresholding problem is employed to demonstrate the proposed approach. The experimental results show that the proper bandwidths are determined automatically and the images are well segmented by the selected thresholds. (C) Elsevier Science Inc. 1997.