The Research Of Sonar Image Segmentation Based On Level-set Method

Sonar image segmentation refers to the sonar image is divided into a series of mutually non-overlapping homogeneous regions, in order to extract the objects of interest. Generally three different intensity regions can be classified as: the echo areas, the shadow areas and the sea-bottom reverberation areas. As a basic technology of image processing, sonar image segmentation plays a vital role on the effectiveness of extracting the feature and recognition rate for objects, such as mines, autonomous underwater vehicle, etc.This paper studies sonar image segmentation methods based on level set models. First, as research on sonar image segmentation methods of home and broad, outlining the applications of the level set methods to image segmentation. Second, this paper introduces the basic theory and principle of level set segmentation and studies the more mature and better segmentation methods of home and abroad, including two-phase level set models of Mumford-Shah model and C-V model, as well as the multiphase level set models of single level set function and multiple level set function. At he same time, an analysis of their strengths and weakness is also given by the experiment results of sonar images segmentation based on these models using matlab in this paper. Finally, this paper proposes new two-phase level set segmentation model and multiphase level set model through joining texture feature based on Gauss-Markov (GMRF), which need not any costly re-initialization procedure, and verifying the models through experiments of sonar image segmentation.