基于MGGD 和上下文模型的小波域图像去噪

考虑到自然图像小波系数分布的两个统计特性，用多元广义高斯分布（MGGD）模型描述小波系数的统计分布，采用基于上下文模型的自适应协方差矩阵估计，再利用最大后验概率估计原始图像的小波系数。实验表明，该方法在直观视觉上去噪效果明显，在信噪比方面也要优于其他的一些小波去噪方法。

Image Denoising Based on MGGD and Context Model in Wavelet Domain

The method uses amodel of Multivariate Generalized Gaussian Distribution（MGGD ）,which takes advantage of tw o statistic characteristic of wavelet coefficients, to describe the statistic distribution of wavelet coefficients of the clean natural im age. The covariance m atrix is estim ated adaptively w ith the method based on context model.The denoised wavelet coefficients are estim ated by m axim um a posteriori（MAP） estim ator. Experim ents show that the proposed method can get a higher PSN R and outperform better than other published methods.