A novel image deblurring method based on high-order MRF prior
Bo Zhao, Wensheng Zhang · 2011
A novel image deblurring method based on high-order non-local range Markov Random Field (NLR-MRF) prior is proposed in the paper. NLR-MRF is an effective statistical framework to model prior knowledge of natural images which leads to excellent performance in some low-level vision problems. In our work, the framework is extended to image deblurring. To overcome some limitations of maximum a-posteriori (MAP) estimation, we adopt Bayesian minimum mean squared error (MMSE) estimation to perform deblurring. The high-order NLR-MRF prior can be easily integrated into this framework. Then, an efficient Gibbs sampling algorithm is employed to compute MMSE estimation. The proposed method frees the user from determining regularization parameter beforehand, which relies on unknown noise level. Our deblurring method shows superior or comparable results to the state-of-art deblurring methods.