Image denoising based on hierarchical Markov random field

We propose a hierarchical Markov random field model-based method for image denoising in this paper. The method employs a Markov random field (MRF) model with three layers. The first layer represents the underlying texture regions. The second layer represents the noise free image. And the third layer is the observed noisy image. Iterated conditional modes (ICM) is used to find the maximum a posteriori (MAP) estimation of the noise free image and texture region field. The experimental results show that the new method can effectively suppress additive noise and restore image details.

Research highlights
► The image region label, clean image and noisy image are represented by a three-layered hierarchical model.
► Each layer is modeled by a Markov random field.
► Iterated conditional modes is used to find the maximum a posteriori estimation of the noise free image and texture region field simultaneously.