Synthetic Aperture Radar (SAR) Image Recovering With Improving Symmetric Alpha-Stable Distribution Method

Synthetic aperture radar (SAR) is a kind of active coherent high resolution imaging radar system capable of producing high-resolution terrain image from data collected by a physically small aperture antenna. Synthetic aperture radar achieve high angular resolution by integrating the backscattered signal to synthesize the effect of a large aperture antenna, and get the high enough range resolution by pulse compressing techniques.However, because of the coherent imaging mechanism, which determining the noise of image is mainly multiply noise. As to the multiply noise, we must convert the multiply noise to additional noise through logarithm transform.Now there are many methods dealing with these problems. But these methods are all based on the Gaussian model in the wavelet domain. As a matter of fact, images wavelet domain band coefficient departures normal distribution often, which causes the performance of these methods dropping severely.SaS (Symmetric alpha-Stable) model is an accurate image model in Wavelet-domain. . But this method needs lots of computation. In this paper, a new model, Gaussian-Cauchy model, was proposed on the base of SaS model. Using this model characterizing the real-world images prior model can reduce computations obviously. In addition, the logarithm transform changes its statistical characteristic, so we must adjust the mean in the process. In this paper, we deal with the statically characters changing particularly after algorithm transform. In the last, we conclude that the method of this paper is feasible form four aspects. The first, using Gaussian-Cauchy method analyzes the SaS model and illuminating Gaussian-Cauchy may depict accurate image model——SaS model. The second,using Gaussian-Cauchy model analyzes wavelet domain coefficient histogram. The third, presenting the experiment results of additional noise (Lena image) and multiply noise (SAR image). The forth, analyzing the computing performance of this method. The experiment results illuminates that the method of this paper may hold on the images edge information and texture characters, in the same time, this method preserves the SAR image structure and may suppress the speckle noise.