Square Distance Symmetric Co-occurrence Matrix Thresholding Method Based on Relative Homogeneity

We studied a novel image thresholding method using symmetric co-occurrence matrix probability information, in which, the relative homogeneity characteristics of the segmented object and background is explored, square distance co-occurrence matrix thresholding criterion is reformed, and the relative criterion of square distance symmetric co-occurrence matrix is proposed. The new thresholding method applies the spatial information of image and takes the relative information between classes into account. The experimental results show that, comparing with the existing correlation methods, the proposed method can extract the object more integrity, and reserve the edge more clearly.