Digital Image Thresholding Segmentation Method Based On Co-occurrence Matrix

Image segmentation is important and fundamental problems in computer vision technology, and a key technology in image processing, analysising and understanding areas.Thresholding method based on co-occurrence matrix is a way of image thresholding segmentation.This paper analyzes the seven existing "gray-gray"symmetric co-occurrence matrix thresholding segmentation methods, compares the performances of seven existing methods through experiments,and shows that the seven existing methods are mainly suitable for simple image,and the segmentation effect of complex image is poorer. In view of this, in order to better use the statistical information of image,this paper defines the means of the object region and background region of the "gray-gray"symmetric co-occurrence matrix,and proposes the thresholding selection method based on square distance and the method based on relative entropy. Simulation results show that the new methods are more adaptable to images and have practical value.