A novel approach for bayesian image denoising using a SGLI prior

Abstract

This paper provides an effective prior for image denoising in a Bayesian framework. The prior combines two well-known discontinuity measures which have been used in illumination normalization methods. We make use of the two measures as a singular new prior for image denoising in a Bayesian framework. Various experiments show that the proposed prior can reduce noise from corrupted images while preserve edge components efficiently. By comparative studies with conventional methods, we demonstrate that the proposed method achieves impressive performance with respect to noise reduction. © 2009 Springer-Verlag Berlin Heidelberg.