Histogram-steered image denoising in the Bayesian framework

Abstract

Rather than concentrating on modeling the image prior probability whose structure is defined locally, in this paper we incorporate the global information from a histogram into the Bayesian method for image de-noising. The key insight is that the histogram of an underlying image can be approximately recovered from the image with additive noise by a deconvolution operation. We test our algorithm in an image set commonly used for denoising test, and obtain improved results.