Bayesian estimation of Two-Sided Gamma random vectors in speckle noise

In this paper, we present a novel speckle removal algorithm within the framework of Bayesian estimation and wavelet analysis. The proposed method to apply a logarithmic transformation to convert speckel, multiplicative, noise model into an additive noise model. The subband decomposition of logarithmically transformed image are the best described by a family of heavy-tailed densities such as Two-Sided Gamma. Then, we propose a maximum a posterior (MAP) estimator assuming Two-Sided Gamma random vectors for each parent-child wavelet coefficients of noise-free log-transformed data and log-normal density for speckle noise. The experimental results show that the proposed method yields good denoising results.