7-5 Training optimization of scalar shrinkage for wavelet color-image denoising

Abstract

This paper addresses the issue of the training optimization of a generic scalar-shrinkage scheme for their application to the wavelet color-image denoising in the shift-inavariant Haar wavelet transform domain. As the generic scalar-shrinkage scheme to be optimized through training, this paper takes up scalar shrinkage utilizing the single-variable shrinkage function, referred to as the Bayesian shrinkage function, which is derived as the Bayesian estimate according a statistical model of the mixture Gaussian distribution.