SAR Image Segmentation Based On Statistical Models

Image segmentation is a key technology in digital image processing. The purpose of image segmentation is to extract the meaningful parts of the image, including the edges and textures, which is the base of image recognition, image analysis and image understanding. In the early research of images, the segmentation methods are mainly two classes, which are the method based on edge and the method based on region, and the method based on statistical models coming up afterwards gets more attention.This paper mainly researches the SAR image segmentation methods based on statistical models both in transform domain and space domain, including the following three aspects:1) We present a SAR image segmentation method based on the second generation Bandelet domain Gaussian Mixture Model (BGMMseg), which not only produces more exact and more continuous edges, but also retain better region information, especially for SAR images which have simple edges and textures.2) We present a SAR image segmentation method based on second generation bandelet-domain HMT-3S model in this paper and the segmentation results not only have more exact and more continuous edges, but also keep better region information. The experiments show that BHMT-3Sseg is efficient and effective for SAR image segmentation.3) We present a SAR image segmentation method based on clustering and Triplet Markov Fields (TMF) in this paper. The experiments show that the segmentation method is efficient and effective especially for the segmentation of SAR images which have complicated backgrounds.