基于形态学成分分析的合成孔径雷达图像去噪  

SAR image despeckling based on morphological component analysis

作　　者：王灿[1] 苏卫民[1] 顾红[1] 邵华[1] 

机构地区：[1]南京理工大学电光学院,江苏南京210094

出　　处：《电波科学学报》2013年第3期448-454,共7页Chinese Journal of Radio Science

基　　金：部基金(9140A13030211BQ02);教育部博士点基金(20113219110018);南京理工大学研究基金(2010ZDJH05;2011ZDJH13);江苏省创新计划(CXLX11_0252)资助课题

摘　　要：合成孔径雷达(Synthetic Aperture Radar,SAR)图像的相干斑抑制一直是SAR图像预处理的重要环节.针对利用小波阈值去噪方法进行相干斑抑制时存在细节丢失的问题,提出一种基于形态学成分分析(Morphological Component Anlysis,MCA)和超完备字典稀疏表示的相干斑抑制方法.该方法使用MCA将图像的平滑部分、纹理部分和边缘部分进行分离,在变换域空间包含脊小波(curvelet)的超完备字典将平滑部分、纹理部分和边缘部分分别进行稀疏表示,相干斑抑制,进行SAR图像的恢复.利用实测SAR图像进行试验,并与Lee滤波、小波阈值等已有方法进行了比较,实验结果表明:本文算法在抑制相干斑的同时更好的保留了有用的细节信息.SAR image despeckling is a prerequisite for many SAR image processing tasks. In view of the problem of detail lost when using wavelet thresholding algo rithm to despeckling, we present a despeckling algorithm based on morphological component anlysis （MCA） and over-complete dictionary sparse representation. In this method, using the MCA theory, we separate the SAR image into piecewise smooth component, texture component and edge component. In the transform do- main space, piecewise smooth component, texture component and edge component are respectively sparsely represented and despeckling by over complete dictionary including curvelet, and then SAR image is recovered out. Real SAR image is used for experiment and we compare our method with Lee filter method, wavelet threshold method and other existing methods. Experimental results show that our algorithm has better despeckling ability and keep more useful detail information of the SAR image.

关 键 词：SAR图像 相干斑 形态学成分分析 脊小波 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]