基于高斯尺度混合模型的小波域图像复原研究  

Image Restoration Based on the Gaussian Scale Mixtures in the Wavelet Domain

作　　者：董湘君[1] 朱鹰屏[1] 

机构地区：[1]广东技术师范学院,广东广州510635

出　　处：《广东技术师范学院学报》2010年第6期23-26,共4页Journal of Guangdong Polytechnic Normal University

摘　　要：将图像分解的小波域金字塔系数的每个邻域模型化为高斯尺度混合模型,并基于该模型提出一种贝叶斯估计方法,复原被线性降晰并叠加一加性高斯白噪声的模糊图像.仿真结果显示,本文算法无论从视觉效果还是峰值信噪比指标上都明显优于维纳滤波复原算法.Each neighborhood of pyramid coefficients of image decomposition in the wavelet domain is mod-eled as the Gaussian scale mixture.A Bayesian estimating method is proposed to restore the image which is blurred by linear distortion and additive white gaussian noise based on the model.The simulation result shows that the restoration effect of the algorithm in this paper is obviously superior to the Wiener filtering restoration whether in the visual effect or in the peak signal-to-noise ratio.

关 键 词：高斯尺度混合模型 小波系数 贝叶斯估计 图像复原 

分 类 号：TN911[电子电信—通信与信息系统]