Image denoising using balanced multiwavelets and scale mixtures of Gaussians
Lahouari Ghouti · 2005
This paper describes a method for image denoising based on a statistical model of the coefficients of balanced multiwavelet transform. The proposed model is inspired by a recent model that has been successfully applied to image denoising using an overcomplete multiscale oriented basis. Clusters of wavelet coefficients are modeled as the product of two independent random variables: a Gaussian vector and a hidden positive scalar multiplier. The existing correlation between the coefficient amplitudes is taken into account by the scalar multiplier. A weighted average of the local linear estimates of the latter represents the Bayesian least squares (BLS) estimate of each wavelet coefficient. Results of denoising images contaminated by additive white Gaussian noise are shown to demonstrate the performance of the proposed method.