SAR Image Despeckling Based on Local Gaussian Model and MAP in NSCT Domain
Xiaoming Bu · Dianzi xuebao · 2010
Applying nonlogarithmic additive model for SAR image,the distribution characteristic of SAR image isotropic regions are researched in Nonsubsampled Contourlet Transform(NSCT) domain,and then Gaussian Model is proposed to model the nonlogarithmic additive noise of the isotropic region in NSCT domain.Based on this model,we apply local sliding windows and adaptively estimate the variation of the nonlogarithmic additive noise,and then the solution of real NSCT coefficient is determined by using Maximum a Posteriori(MAP).Without applying logarithmic transform for original SAR image,the proposed method holds a good ability of radiometric preservation,the speckle is despeckled well while has less Pseudo-Gibbs phenomenon in uniform regions,edges and textures of despeckled image are also clear.Our algorithm possesses high performance over many traditional algorithms according to vision or objective evaluation.