MRF and CRF Based Image Denoising and Segmentation

Abstract

In this work, we employ a pair wise Markov Random Field (MRF) and a Conditional Random Field (CRF) for bi-level image segmentation and denoising. For both tasks, the Ising pair wise model and the Iterative Conditional Mode (ICM) inference method are implemented, assuming the parameters of the unary and pair wise potentials are known. Experimental results demonstrate the effectiveness of the proposed algorithm.