The Speckle Suppression Algorithms Based On Coupling Similarity And Reinforcement Of Iteration For SAR Imagery

Synthetic Aperture Radar(SAR)has been widely used in many fields because of its ability to capture information on earth surface in both day and night.However,due to the speckle characteristics,it is very difficult to interpret the Synthetic Aperture Radar(SAR)images.In recent years,in order to suppress the influence of SAR image speckle,there are many speckle suppression algorithms,among which the nonlocal means and anisotropic diffusion algorithms playing an important role in speckle suppression.However,two problems still could be seen in the nonlocal means algorithms on SAR images: one is the differences on distances with different high scattering coefficients demostrating less discrepancy.The other is he attenuation coefficient lacking automaticity.While anisotropic diffusion algorithms show over-smooth on edge and texture,together with the number of iterations selected empirically.In order to alliviate these problems,this thesis focus on nonlocal means and anisotropic diffusion algorithms to suppress speckly in SAR imagery.The main work of this thesis is as follows:1)A nonlocal mean speckle suppression algorithm based on the similarity coupling of region and pixel is proposed.First,the similarity differences are little among different adjacent homogeneous regions with high backscattering coefficient,causing dissimilar region aggravating the smoothness on current homogeneous region.Some existing algorithms used to introduce auxiliary terms to improve the algorithm performance.Different from these algorithms,this thesis presents a similarity measure based on the coupling of region and pixel.It has been proved theoretically that the similarity distance is available for multiplicative model,and also verified that the similarity distance is superior to those of existing algorithms.Then,a kind of adaptive attenuation coefficient on gamma-like function is given to solve the problem of non-automaticity of attenuation coefficient in existing algorithms.Finally,a discriminating strategy for the size of similar blocks is given.In the experiments,two synthetic speckled images and five real SAR images with different resolutions are used for evaluation and analysis.The experimental results show that the proposed algorithm has better suppression performance than the existing speckle suppression algorithms,and the edge retention performance is close to the best performance of the comparison algorithms,while the mean ratio and point target retention performance demostrate good effects.It is shown that the similarity measure and adaptive attenuation coefficient given by the proposed algorithm are feasible and solve the problems in the existing non-local means algorithms.2)An anisotropic diffusion speckle suppression algorithm based on enhanced diffusion coefficient on iteration is presented.To tackle the problem that edges and textures could be oversmoothed during iterations in existing anisotropic diffusion algorithms,a diffusion coefficient driven by edge mapping strength,iteration threshold and iteration number is proposed.This diffusion coefficient has advantages that as the number of iterations increases,the algorithm gradually does not smooth the edges and textures,and maintains despeckling behaviour in homogeneous regions.Moreover,an iterative adaptive step-size strategy is given to further improve the ability of smoothness in homogeneous regions and preservation in edge and texture.Furthermore,an iteration termination condition is presented on the ratio image.The experiments are verified by the synthetic image and the actual images with five resolutions.The experimental results show that the proposed algorithm is better than the contrast anisotropic diffusion algorithms in smoothing and edge retention.It has demonstrated good results in other performance evaluation,indicating that this algorithm could solve the problem in existing anisotropic diffusion algorithms.