Fuzzy Partition and Correlation for Image Segmentation with Differential Evolution

Thresholding-based techniques have been widely used in image segmentation. The selection of appropriate threshold is a very significant issue for image thresholding. In this paper, a new image histogram thresholding method based on fuzzy partition and maximum correlation criterion is presented. In the proposed approach, the regions, i.e. object and background, are considered ambiguous in nature, and hence the regions are transformed into fuzzy domain with membership functions. Then, the fuzzy correlations about regions are constructed and the optimal threshold is determined by searching an optimal parameter combination of the membership functions such that the correlation of the fuzzy partitions is maximized. Since the exhaustive search for all fuzzy parameter combinations is too costly, the differential evolution algorithm is introduced into fuzzy correlation image segmentation to solve this optimal problem adaptively. Experimental results on general images and infrared images demonstrate the effectiveness of the proposed method.