Research On SAR Image Segmentation Based On Spatial Domain MRF Model

The interpretation of Synthetic Aperture Radar(SAR)image has become the most important part of the research of SAR.As the initial step of interpretation,SAR image segmentation has also become a hot research topic.The algorithm of SAR image segmentation based on Markov Random Field(MRF)model,according to its taking full advantage of images’ statitical properties and local features,which is good to SAR image segmentation,has been taken seriously by researchers.However,due to the complexity of optimizition,the speed of acquiring good segmentation is pretty slow.It is necessary to speed up the traditional SAR image segmentation algorithm based on MRF model.The speed of SAR image segmentation shall be improved based on the main idea of reducing to-be-processed objectives in spatial domain MRF model.Then a revised algorithm is proposed based on superpixel algorithm and MRF-model-based SAR image segmentation algorithm.The contents include:1.Image segmentation algorithm suitable for SAR image based on MRF model is researched.And FGMM and Potts model is adopted to build MRF model for SAR image based on the principle of MRF in image segmentation.The above model will be used into SAR image segmentation combined with ICM segmentation method.2.Suitable superpixel algorithm is selected based on SAR image’s traits and the principle of MRF model in image segmentation.After analysising the differences and similarities of superpixel algorithm’s preprocessing on image,considering the result of preprocessing’s situation of neighborhood and pixel pack’s accuracy,SMRF(Superpixel-MRF)algorithm will be researched.And the analog experiment analog demonstrates the promotion of speed and quality assurance of SMRF algorithm.3.The neighborhood system generation aim to the question of quality decrease of SMRF is researched.Pseudo-meshing and weighting-based neighborhood algorithm will be put forword from meshing-based and weighting-based angles.Improved method has higher quality of image segmentation while keep above algorithm’s advantage.By designing experiment analog,the thesis will demonstrate the new neighborhood system can bring SMRF satisfied speed and quality of SAR image segmentation.