Bayesian wavelet-based image denoising using Markov Random Field models
Yanqiu Cui, Tao Zhang, Shuang Xu, Houjie Li · 2010
This paper presents a Bayesian denoising method based on Markov Random Field (MRF) models in wavelet domain in order to improve the image denoising performance and reduce the computational complexity. The computations of the initial mask, optimal mask and shrinkage factor of the wavelet coefficient are the core of this method. To obtain the appropriate initial mask, a simple two-state Gaussian mixture model is constructed and an estimation method of the initial mask based on the maximum a posteriori (MAP) criterion is proposed. Based on this initial mask, an optimal mask is obtained. To reduce the computational complexity of the optimal mask, a simple optimization method, the iterated conditional modes (ICM) method is adopted. A Bayesian wavelet shrinkage factor is derived based on this optical mask. Under this framework, the computational complexity of the denoising method can be reduced. Simulation results demonstrate our proposed method has a good denoising performance while reducing the computational complexity.