Research On Remote Sensing Image Segmentation Based On Edge Detection And Markov Random Field Model

Image segmentation is an important process of the image project. It is the only way from image processing to image explanation and image analysis. People began to study image segmentation long before because of its importance. Up to now, many theories and methods have been proposed and applied to image segmentation. In the remote sensing field, because the RS images have the characters of big size, multi spectrum, various content, sufficient texture feature, multi scale etc, apply usual image segmentation methods to RS images can't give satisfying results. It is necessary to study image segmentation methods applicable to RS images. This paper used RS images as the primary segmentation object, made a study of RS image segmentation methods, and proposed an effective method for RS image segmentation.The research of the paper includes:(1)The segmentation methods and RS segmentation methods of these years are reviewed, the advantages and disadvantages are analyzed, and some potential research fields are pointed out.(2)LIP model resolved the out-of-range problem of traditional image processing model , so it is more suitable for various processing of images.This paper proposed a LIP model based edge detection method.the simulation experiments are presented to show the efficiency of the method and its advantage over the traditional edge detection operators and the LIP-Sobel operator.And the character that the methods based on the LIP model are sensitive for lower intensity images is proved theoreti-cally.(3)In order to resolve the local optimum problem of CLARANS,a Simulated Annealing based CLARANS clustering method is proposed. (4)A edge detection based initial segmentation method is proposed, GMRF model is used for texture feature extraction ,and the simulated Annealing based CLARANS clustering method is used for segmentation of RS images.Finally.the research of the paper is summarized and future research is expected.