中文摘要：

建立了广义全变分（total variation,TV）模型,分析正则项在复原算法中的作用.分别从图像的平坦区域和边缘区域入手,在平坦区域图像各向同性扩散,在边缘区域则要满足各向异性扩散,从理论上对两种情形下的扩散做深入分析,推导出广义TV模型满足的一些条件,为了防止高噪声情形下复原模型失效以及克服方块效应,在正则项中引入了Contourlet收缩,它是一种多分辨的、局域的、多方向的更稀疏的图像表示方法,正则项中引入的Contourlet收缩具有去噪和提取图像重要信息的作用,Contourlet收缩与广义TV正则化相结合,兼顾了图像的光滑性和边缘保持,特别是在图像严重模糊、噪声越多的情形下,更加体现了这种算法比改进的TV模型有效.

英文摘要：

A generalized total variation（TV） model is established,and the key role of regular item in restoration algorithm is analysed from aspects of the flat areas and the edge region image respectively.Image needs to be an isotropic spread in the flat areas and be the anisotropic diffusion in the edge region.This paper derives some conditions of general TV model that need to be satisfied based on the theoretical analysis of diffusion in the flat areas and the edge region.The new method of introducing Contourlet shrinkage to the regularization is provided in order to prevent recovery model failure in highnoise cases and overcome the effects of the blocking artifacts.Contourlet is multiresolution,local,multi-direction image sparse representation.Contourlet shrinkage that is introduced to the regularization item plays the role of the de-noising and extracting the important information from image.At last,Experiment results show that the new combination method of Contourlet shrinkage and generalized TV regularization takes into account the image smoothness and edge-preserving. Especially,in the image severely blurred and the more noise cases,the new model is efficient than the improved TV models.