Despeckle Of Polarimetric SAR Based On Non-local Means

Compared with the single polarized SAR(synthetic aperture radar),polarimetric SAR(Pol-SAR) includes more information of targets and polarimetric,which is one of the important technology of remote sensing applications. Because of the coherent imaging properties of Pol-SAR, the further application of terrain classifiction, target detection and recognition is affected seriously by the internal speckle noise. Therefore, despeckling becomes the foundation step in Pol-SAR information processing. In this paper, a research and the exploration of Pol-SAR Speckle Noise Model and Despeckle is made, we proposed speckle noise suppression algorithm for Pol-SAR data based on the new speckle model of Pol-SAR. The main tasks of this article is as follows:(1)The impact of similar block shape of non-local mean filtering is analyed. we proposed speckle noise suppression algorithm the shape adaptive Non-local mean filtering method for Pol-SAR data based on the multiply-add speckle model.of Pol-of SAR. The experiment shows that this method can effectively remove speckle the same time effective to maintain the edge details;(2) Analysis of the non-local mean filtering the search window memory associated with the current pixel is not the pixels, and then use the LPA-ICI technique to estimate the anisotropy of the pixel neighborhood in order to remove irrelevant pixels, and finally made the anisotropic non-local mean despeckleing method for Pol-SAR. The experiment shows that the filtering effect of the method was significantly better than the classic refined Lee filter and improved sigma filter;(3) Proposed a singular value decomposition (SVD) method of non-local mean despeckleing for Pol-SAR data. The method is maked by the span matrix singular value decomposition to obtain the feature information, which is used to measure the similarity between pixels. The experiment shows that this method can effectively remove speckle, more smooth homogeneous regions, while maintaining the edge details, and better visual effects.This research is supported by the National Natural Science Foundation of China (No:61173092),the Fundamental Research Funds for the Central Universities (No.JY10000902045) and the Fund for Foreign Scholars in University Research and Teaching Programs (the111Project)(No:B07048).