Maximum a posteriori Image Denoising with Edge-preserving Markov Random Field Regularization

Image denoising is still a challenging problem for researchers. The conventional methods for image denoising often smooth the images and result in blurring the edges. In this paper, however, we consider not only the noise removal ability but also another important requirement for image denoising procedures that is true image structures, such as edges, should be preserved. To this end, we propose a maximum a posterior (MAP) image denoising algorithm using a novel edge-preserving Markov random field (MRF) model. Considering edges tend to be continuous in space, the connectivity of structure tensor is defined to describe edges. And the revised Markov random field model is proposed by using the edge connectivity, which can adaptively control the degree of smoothing. The experiments show that our new method gives superior performance in terms of both objective criteria and subjective human vision when compared with related MRF models.