关键词: 图像去噪, 马尔可夫链蒙特卡洛方法（MCMC）, 方差估计, 预处理, 并行处理

Abstract: Image denoising is a prerequisite for the many processing tasks of image. Markov Chain Monte Carlo algorithm is an important method of image denoising. However, the method has some problems such that the denoised image has obvious spots, the denoising image corrupted by heavy noise is not satisfactory, the noise variance needs to be estimated in practical application, and the operation speed of this method is slow. This paper proposes a two-step denoising method. It preprocesses the noise image using the mean filter. It estimates the pretreated image noise variance. It uses the MCMC image denoising method. To take full advantage of multi-core processor resources, this paper studies the parallel programming of MCMC algorithm. The multi-core program increases the speed of MCMC algorithm. The experiments show that the denoising method given in this paper reduces spots and improves the signal-to-noise ratio. Parallel processing can make the algorithm more efficient.

Key words: image denoising, Markov Chain Monte Carlo（MCMC）, variance estimation, pretreatment, parallel processing