SAR image denoising based on alpha-stable distribution and Bayesian wavelet shrinkage
Xin Xu, Yin Zhao, Wanbin Zhou, Yijin Peng · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2009
In this paper, an algorithm for synthetic aperture radar (SAR) image denoising in the wavelet domain is presented. The alpha-stable distribution is applied to model the wavelet coefficients of the logarithmically transformed SAR images and the Gaussian mixture model to represent the Speckle. The method of regression-type is used to estimate the four parameters of the alpha-stable distribution and EM algorithm to estimate the variance of the noise respectively. Since the alpha-stable distribution do not always have a closed-form formula, Zolotarev's (M) parameterization is exploited to obtain the probability density function (PDF) of the alpha-stable distribution. Consequently, a maximum a posteriori (MAP) estimator is designed based on the alpha-stable prior to restore the SAR image. The experimental results, including simulated SAR image and SIR-C/X-band SAR image, indicate that the proposed algorithm has capability both in Speckle suppression and details preservation.