SURE准则的非局部SAR图像相干斑抑制

发布时间：2018-02-14 15:06

本文关键词： 各向异性高斯窗非局部均值 SAR图像降斑 Stein无偏风险估计　出处：《电子科技大学学报》2014年01期 　论文类型：期刊论文

【摘要】：针对传统空域非局部平均方法在合成孔径雷达图像相干斑抑制中存在相似区域提取和方向信息捕获不足的问题,提出了一种基于各向异性高斯方向窗和Stein’s无偏风险估计(SURE)准则融合的非局部均值(NLM)算法。该方法设计多个不同方向的各向异性高斯窗来匹配SAR图像的局部空间几何结构,比传统的方形窗能更好地保护SAR图像中的方向性结构。采用比率测度来衡量图像块间的相似程度,并计算基于该各向异性高斯窗的NLM结果。结合SURE准则来融合不同方向的各向异性高斯窗的非局部平均结果,获得最终的SAR图像降斑结果。针对多幅SAR图像进行对比实验,实验结果表明:该方法在有效抑制SAR图像相干斑的同时能很好地保留图像的几何结构信息,为后续的SAR图像理解与解译提供了良好的基础。
[Abstract]:In order to solve the problem of similar region extraction and direction information acquisition in SAR image speckle suppression based on traditional spatial nonlocal averaging method, a new method is proposed. A nonlocal mean value (NLM) algorithm based on the fusion of anisotropic Gao Si directional window and Stein's unbiased risk estimation criterion is proposed. In this algorithm, several anisotropic Gao Si windows in different directions are designed to match the local spatial geometric structure of SAR images. Compared with the traditional square window, it can better protect the directional structure in the SAR image. The ratio measure is used to measure the similarity between the image blocks. The NLM results based on the anisotropic Gao Si window are calculated. Combining with the SURE criterion, the nonlocal average results of the anisotropic Gao Si window in different directions are fused to obtain the final SAR image speckle reduction results. Experimental results show that the proposed method can effectively suppress the speckle of SAR images while preserving the geometric structure information of the images, which provides a good basis for the subsequent interpretation and interpretation of SAR images.
【作者单位】：西安电子科技大学智能感知与图像理解教育部重点实验室;
【基金】：国家自然科学基金(61173093,61072106,61003198,61001206) 教育部长江学者和创新团队支持计划(IRT1170)
【分类号】：TN958

【参考文献】