小波域非参数贝叶斯估计图像去噪  

Novel Nonparametric Bayesian Estimator for Image Denoising in Wavelet Domain

作　　者：邓承志[1] 曹汉强[1] 汪胜前[2] 

机构地区：[1]华中科技大学电子与信息工程系,湖北武汉430074 [2]江西科技师范学院江西省光电子与通信重点实验室,江西南昌330013

出　　处：《小型微型计算机系统》2008年第10期1855-1859,共5页Journal of Chinese Computer Systems

基　　金：国家自然科学基金(60462003,60772091)资助

摘　　要：提出一类新的非参数贝叶斯估计器,估计器利用正态反高斯(NIG)分布作为先验模型.与广义高斯分布(GGD)、alpha稳定分布和贝塞尔K分布(BKF)相比,正态反高斯分布能更加精确地对图像小波系数分布进行拟合.在二次型贝叶斯规则下,推导出基于正态反高斯分布的后验条件均值估计.最后,对图像进行去噪实验.实验结果表明,与最新提出的算法相比,该方法获得更高的峰值信噪比增益和好的视觉效果.A novel nonparametric Bayesian estimator for image denoising in wavelet domain is presented. In this approach, norreal inverse Gaussian （NIG） distribution is used as a prior model to capture the sparseness of the wavelet expansion. Compared with other distributions, such as the generalized Gaussian distribution （GGD）,α-stable models, and Bessel K forms （BKF）, it fits very well to the distributions of wavelet coefficients of natural images. Based on Lz based Bayes rules, a posterior conditional means estimator is designed. Finally, the estimator is used to image denoising. Experimental results show that compared with several recently published algorithms, the proposed method achieves state-of-art performance in terms of peak signal-to-noise ratio and visual effect.

关 键 词：正态反高斯分布 小波变换 后验条件均值估计 图像去噪 

分 类 号：TP911.73[自动化与计算机技术]