Study On Methods For Speckle Reduction And Segmentation Of SAR Image

On the basis of previous studies of many pioneers in this field, some investigationsare done in this paper as follows: First, fundamentals of the Synthetic Aperture Radar(SAR) and the characteristics of SAR image are introduced, and after that, the specklein the SAR images has been reduced to a great extent successfully using Maximumlikelihood (ML) rule and the Maximum a posterior probability (MAP) rule. In addition,the efficiencies of the several speckle removal methods are compared and analyzed.Second,SAR image is statistically analyzed by using the circular kernel vectors ofgray-level co-occurrence matrices(GLCM). As the radius of the circle increasing, thecircular kernel is biased toward the response from feature on the periphery. Tocounteract this problem the GLCM statistics can be weighted as a function of radius.And a variety of textural measures, which can reflect the characters of SAR imagetexture, can be calculated by the GLCM statistics, such as inverse difference moments,entropy, second rank moments and contrast, etc. The real contour can be recoveredwhile the speckle in the SAR image can also be removed efficiently. With differentparameters, different disposing results can be achieved from the original SAR images.Experimental results show that this method is practicable and applicable. Last, the SARimage segmentation has been studied using simulated annealing (SA) method withMarkov correlated structured neighborhood Models under MAP rule and the satisfiedresults has been seen in the simulation.