Bilateral signal variance estimation for wavelet-domain image denoising

Abstract

The estimation of the signal variance is a critical challenge in wavelet-domain minimum mean square error (MMSE) based image denoising. In contrast to the conventional approaches that treat the neighboring wavelet coefficients equally to estimate the signal variance at each coefficient position, here an adaptive approach is proposed that utilizes a bilateral statistical scheme adaptively adjusting the contributions of neighboring wavelet coefficients to provide an accurate estimation of the signal variance. Experimental results are presented to demonstrate the superior performance of the proposed approach.