基于异质性测量和非局部平均的斑点噪声抑制  

Novel speckle filtering based on heterogeneity measurement and non-local mean

作　　者：陈少波[1,2] 侯建华[1,2] 熊承义[1,2] 张华[1,2] 

机构地区：[1]中南民族大学电子信息工程学院,武汉430074 [2]中南民族大学智能无线通信湖北省重点实验室,武汉430074

出　　处：《计算机应用研究》2014年第8期2557-2560,共4页Application Research of Computers

基　　金：国家自然科学基金资助项目(61201448);湖北省自然科学基金资助项目(2011CHB043;2012FFA113);中南民族大学中央高校基本科研业务费专项资金自科一般项目(CZY10001)

摘　　要：传统的欧氏距离不能鲁棒地度量含有相干斑噪声的SAR图像块之间的相似性。针对这一问题,将SAR图像的异质性测量方法与传统的欧氏距离结合,产生了一种新的SAR图像相似性测度,并在此基础上,提出了一种新的用于斑点噪声抑制的非局部平均滤波算法。该算法首先计算相似性窗口之间的欧氏距离和搜索窗口的变差系数;然后利用搜索窗口的变差系数自动调整退化参数h,并在欧氏距离和调整后的退化参数的基础上计算新的SAR图像相似性测度;最后利用新的相似性测度对待处理像素点进行非局部平均恢复。对仿真与实际SAR图像的斑点噪声抑制实验表明,新算法能有效去除斑点噪声和保留边缘纹理等细节区域,视觉效果很好;且与现有的非局部抑斑算法相比,计算复杂度大大降低。Conventional Euclidean distance can not measure the similarity of SAR image patches robustly. In order to solve this problem,the Euclidean distance was combined with SAR image heterogeneity measurement. Then,this paper proposed a novel similarity measure of SAR image. Further,it designed a new speckle reducing algorithm based on the novel similarity measure. First,it calculated the Euclidean distance of similarity windows and CV（ coefficient variation） of search window.Secondly,it adjusted the decay parameter h to CV. Then,it calculated the novel similarity based on the Euclidean distance and the adjusted decay parameter. Last,it performed a weighted average of the values of similar pixels based on the new similarity. The deal with synthetic and real SAR images contaminated by speckle would be filtered by this proposed algorithm. The visual quality and the quantification estimation show that the proposed approach can suppress speckle effectively and keep features of edge,texture,and details simultaneously. In addition,comparing to other NLM methods for speckle filtering,the computational complexity of the proposed is greatly reduced.

关 键 词：SAR图像 相干斑 非局部均值 异质性测量 滤波 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]