Nonlocal Means Speckle Suppression Algorithms Driven By A Q-characteristic Decay Coefficient And Monoblock Similarity

Synthetic Aperture Radar(SAR)imaging has been widely used in the monitoring of land covers,marine,or the environment,due to its all-time acquisition capability.However,the inherent speckle seriously affects the subsequent processing of SAR image.Therefore,in order to reduce the impact of speckle on SAR image,many speckle suppression algorithms could be seen in recent years,among which the nonlocal mean algorithm is one of hotspots.However,there are still two problems in the nonlocal mean algorithms for SAR images: 1)Decay coefficient could not reach a good balance between speckle suppression and edge preservation;2)Computational burden still exists.In order to solve the above problems,this thesis studies the nonlocal mean algorithm of SAR image from the calculation of decay coefficient and block similarity.The main work of this thesis is as follows:(1)A nonlocal mean speckle suppression algorithm based on Q characteristic decay coefficient is proposed.For the problem to balance between speckle suppression and edge,an exponential decay coefficient based on Q feature is presented.The decay coefficient could be used automatically to calculate the decay coefficient of the current region according to the eigenvalues of each region.Its purpose is to make the edge as small as possible and the homogeneous region as large as possible,so as to achieve the balance between homogeneous and nonhomogeneous regions.In order to evaluate the decay coefficient,it is necessary to use the similarity measure its effectivemess.Therefore,tje theoretical analysis on the likelihood similarity commonly used in nonlocal mean algorithm has been exploited.Through the theoretical analysis in heterogeneous region,it is found that there is the phenomenon of over smoothing or under smoothing.In order to solve this problem,a correction term based on local statistics is introduced into likelihood similarity to correct the distance.Therefore,the rationality of the correction is analyzed theoretically,and the effectiveness of the correction is analyzed by experiments.In order to verify the performance of the algorithm,synthetic and actual SAR images are used in experiments.The experimental results show that the proposed algorithm could generally be better than the comparison algorithms in speckle suppression and edge preservation,which verifies the effectiveness of the proposed algorithm.(2)A Block matching nonlocal mean speckle suppression algorithm based on nonlinear mapping is provided.Aiming at the problem of computational burden in the existing nonlocal mean methods,this thesis presents the idea that only one pair of blocks could be used to calculate similarity in the searching window.The similarity of other blocks could be obtained through the similarity of above one pair of blocks.First,the nonlinear relationship between the similarity measures of different adjacent blocks and central blocks in homogeneous and heterogeneous regions is theoretically derived.The theoretical results show that the similarity measures of multiple pairs of blocks in the searching window could be expressed by a pair of block similarity with nonlinear mapping,and the look-up table is calculated by the coefficient of the nonlinear expression,and then the similarity distance table is calculated by the look-up table,so as to reduce the computational complexity.In order to verify the performance of the algorithm,synthetic and real SAR images have been used for experiments.The experimental results show that the speckle suppression ability of the proposed algorithm has a good performance in contrast to some existing algorithms.At the same time,the running time is analysized after establishment of look-up table.The average running time on the tested data set is 6.88406 s,which has certain advantages compared with other algorithms.However,the time complexity to establish look-up table is a bit time consuming,which is required with further study.