结合多尺度边缘检测的SAR结构邻域滤波方法

噪声抑制是合成孔径雷达（SAR）图像处理的一个重要环节，通常的方法在抑制噪声的同时也使得图像的边缘模糊。该文提出一种结合宏观边缘信息的SAR图像结构邻域滤波方法。首先利用多尺度边缘检测算法获得SAR图像主要的边缘信息，在此基础上对结构邻域滤波法进行改进。结构邻域滤波法用一系列反映图像局部方向信息的邻域模板描述图像的细节特性，滤波过程中引入图像的宏观边缘信息，对邻域模板的选择范围进行约束。最后运用模拟退火算法选取合适邻域模板对目标点的强度进行MAP估计。实验表明该方法能够较好地保持图像的边缘特征，同时有效抑制了斑点噪声。

A SAR Image Filter Method Based on Structured Neibourhood Model with Multi-scale Edges Detection

Speckle reducing is an important step in Synthetic Aperture Radar (SAR) image processing, ordinary methods blur image edges while reducing speckle. A SAR image filter method based on structured neibourhood model with macroscopical edge information is proposed in this paper. First multi-scale edge detector is applied to achieve principal image edges, then structured neibourhood filter is modified and used. Structured neibourhood filter adopts a series of templates reflecting regional direction information to describe image detailed property, the macroscopical edge information is introduced to constrain the selection of templates while processing. Finally simulated annealing algorithm is used to get appropriate template, the MAP estimate of the object pixel intensity can be achieved. Experiment indicates this method reduces effectively speckle while preserving image edges.