关键词: 合成孔径雷达（SAR）, 相干斑抑制, 影像直方图, 统计分布函数, 异常像素检测, 统计特性保持

Abstract:

The speckle phenomenon in synthetic aperture radar (SAR) images is caused by the mutual interference of backscattering signals of ground objects. Although it is similar to noise, it involves the scattering characteristics of ground objects. The scattering characteristics of the same ground object can be described by the speckle statistical distribution model. Therefore, in the process of noise reduction, the scattering information of the ground object contained in the speckle can be preserved by restoring the statistical distribution characteristics of speckle. Based on this idea, an iterative filtering method of SAR image based on speckle statistical characteristic preservation is proposed. The proposed method assumes that the statistical distribution function of a given SAR image is known a priori, namely the mixture Gamma distribution, and its distribution parameters can be estimated by the pixel values of the image. Then, the EM（Expectation Maximization） algorithm is used to segment the SAR image based on the mixture Gamma distribution, so as to obtain the homogeneous regions in the image. Subsequently, for different homogeneous regions, select the gray levels with larger fitting errors, calculate the densities of the pixels whose values are the gray levels according to the segmentation results, and judge whether they are abnormal pixels, and the abnormal pixels are filtered by Frost filter. Repeat above procedure until the histogram of filtered image fits the statistical distribution function well. Experimental results of GF-3 and Radarsat-2 SAR image show that, on the premise of maintaining image quality, the proposed method can not only obtain better statistical modeling results, but also suppress speckle well and achieve image denoising.

Key words: Synthetic Aperture Radar (SAR), speckle reduction, image histogram, statistical distribution function, abnormal pixels detection, statistical characteristic preservation