徐新, 廖明生, 朱攀, 卜方玲. 单视数SAR图像Speckle滤波方法的研究[J]. 武汉大学学报 ( 信息科学版), 1999, 24(4): 312-316.
引用本文: 徐新, 廖明生, 朱攀, 卜方玲. 单视数SAR图像Speckle滤波方法的研究[J]. 武汉大学学报 ( 信息科学版), 1999, 24(4): 312-316.
Xu Xin, Liao Mingsheng, Zhu Pan, Pu Fangling. Research on Speckle Filtering of Single-look SAR Image[J]. Geomatics and Information Science of Wuhan University, 1999, 24(4): 312-316.
Citation: Xu Xin, Liao Mingsheng, Zhu Pan, Pu Fangling. Research on Speckle Filtering of Single-look SAR Image[J]. Geomatics and Information Science of Wuhan University, 1999, 24(4): 312-316.

单视数SAR图像Speckle滤波方法的研究

Research on Speckle Filtering of Single-look SAR Image

  • 摘要: 在分析单机数SAR影像的Speckle局部统计特性和现有的空间滤波算法的基础上,提出了一种新的Speckle滤波方法,该方法采用特殊的滤波窗口邻域划分方法,根据窗口相对标准差自适应地调整窗口尺寸和窗口内的滤波区域。该方法被用于多幅单视数ERS-1/2SAR图像去Speckle处理,并与以往的典型算法进行了比较,取得了较好的实验结果。

     

    Abstract: Since synthetic aperture radar (SAR) is a coherent system, images acquired by SAR are accompanied with speckle noise which disturbs image interpretation and target classification. This noise is conspicuous in single-look SAR images. Based on the analysis of the local speclke statistics characteristics of single-look SAR image and algorithms of spatial filtering,an adaptive speckle filtering method is presented in this paper. The Kuan filter which is based on MMSE criterion is modefied in the method. One of the focal points of the algorithm is to find the set of pixels which belong to the same homogeneous area within filtering windows adaptively. The local relative standard deviation within filtering windows is und as the crucial parameter while a special neighborhood model is used in the filtering window in order to choose the largest homogeneous area within filtering windows. Firstly the formal relative standard deviation Cx of filtering window is calculated.Then the filtering window is divided into eight mutually exclusive sub-window and the central pixel is the only repetitious pixel in each sub-window. For every sub-window, the local relative standard deviation Ci is calculated. If Cx is large than a threshold Cu,the sub-window with the greatest Ci is removed from filtering window and Cx of remaining pixels in filtering window is recalculated. Repeat the procedure until Cx<Cu. The remaining sub-window are meed as filtering area. The selected filtering area is the largest homogeneous area in filtering window and the Kuan filtering algorithm is adopted within the filtering area. If all sub-window within filtering window are removed, reduce the size of window for the determination of homogeneous area. If the size of window is 3 by 3 pixels, replace the central pixel with the average of the nearest four pixels around it. The method is applied to several single-look ER-1/2 SAR images. The results show that the performance of the approach presented is satisfactory in both speckle filtering and the Preservation of image details, and in generating visually-natural images as well.

     

/

返回文章
返回