A new bivariate model with log-normal density prior for local variance estimation in AWGN

This paper concerns wavelet-based image denoising using Bayesian technique. The conventional probability density function (PDF) used in the denoising process usually has parameters that are calculated from the first few moments only (for example, mean and variance). In this work, a new image denoising algorithm based on bivariate Pearson Type VII distribution is presented. This PDF is used in view of the fact that it allows higher order moments to be incorporated in the probabilistic modeling of the wavelet coefficients. One of the cruxes of the Bayesian image denoising methods is to estimate statistical parameters for shrinkage function. In this paper, maximum a posteriori (MAP) is used for local variances with Log-normal density prior for local observed variances and Gaussian distribution for noisy wavelet coefficients. The experimental results show that the proposed method gives good denoising results.