Thresholding approaches with interval-valued fuzzy sets to image segmentation

Thresholding approaches are fundamental and important techniques in image segmentation. There exist lots of integrated techniques of thresholding methods with fuzzy sets theory. The fuzzy compactness and fuzzy divergence are powerful tools in dealing with vague and imprecise data, and are widely applied to thresholding segmentation for images. This paper extends the two kinds of fuzzy measures to interval-valued fuzzy sets to eliminate uncertain assignments of membership degrees of pixels in images. Two thresholding techniques are proposed for image segmentation. The affinity characteristics of pixels in images are sufficiently considered in the new techniques. The experimental results show that the selection of initial membership functions brings little impact on thresholding segmentation of images.