Sar Amplitude Images Speckle Reduction Of Variational Pde Model And Algorithm Research

Since the emergence of the SAR (Synthetic Aperture Radar), it has been widely applied in military, civilian and other fields, for example: deep space exploration, earth remote sensing and imaging to detect. The Image resolution of the SAR includes spatial resolution and radiometric, which is important for understanding, analysis and pattern recognition about SAR image. The speckle about SAR exerts severe impact on the image quality. Thus, how to curb the speckle has always been a hot spot of research.This paper starts from the SAR imaging mechanism, focuses on the model of variational partial differential equations, and combines Bayesian method and Regularization method. The aim is for researching theory and algorithm for removing the noise of image effectively. The main job and the great achievements it has made are as follows:Firstly, it makes the research on classic filtering algorithm of SAR, about local statistical properties of algorithm based SAR (Lee filter, Kuan filter, Frost filter, Sigma filter and exquisite gamma map filter). Also experiments have been done in order to analyze the merits and demerits of classic filtering algorithm.Secondly, it researches and achieves Aujol and Aubert's image restoration model and Huang's fast algorithm. Also it has discovered the details of maintaining image concerning these two patterns. Based on the theory of non-local regularizing, this thesis has proposed a refreshed SAR pattern based on non-local regularizing and texture fidelity. This pattern has made improvements in the part of regularizing and a new and regularizing priori model has been put forward. Since regularizing priori model has respect to the weighting function, three different weighting functions have been advanced after the research of relations between the geometric distance of pixels and the similarity. Also it researches effect of removing the noise of image using three different weighting functions. For the new pattern integrates AA and the ideas of non-local regularizing, it can retain the details of image better while remove the noise of image effectively. And experiments have verified the above point.Thirdly, it has proposed a new algorithm: nonlinear diffusion algorithm of dual-tree Complex wavelet - TV, coupling Soft-threshold method of dual-tree complex wavelet and Aujol and Aubert's regularizing method. The new algorithm can remove the noise of image effectively using experiments. The merit of this algorithm can be confirmed by juxtaposing Aujol and Aubert's image restoration model and Huang's fast algorithm.