A new SAR image filter for preserving speckle statistical distribution

Abstract

In general, the existing spacial filters occupy themselves in smoothing out the speckle noises pixel by pixel in a local or non-local way, without preserving the information hiding in the speckle noises. This paper presents a new filtering paradigm that aims to restore the statistical characteristic of speckle distribution, rather than simply smoothing them. Firstly, under the assumption that the statistical distribution function corresponding to the homogeneous regions in a given noising SAR image is known a priori, and their distribution parameters can be estimated with the values of pixels in these regions. Several gray levels with large deviation are selected by comparing the histogram and the statistical distribution function. Then, the pixels taking these gray levels as their values can be spatially located in the image domain, some of which are considered as abnormal pixels after abnormal pixels detection based on ratio pixel relativity measure, and these abnormal pixels are to be filtered. Finally, repeat the above procedure until the histogram of filtered image fits the distribution function well. Tests on real SAR images show that the proposed method can obtain better statistical modeling results, and simultaneously achieve noise suppression to some extent while maintaining image quality better.