交替方向乘子法求解混合全变差模糊图像复原模型  

Alternating Direction Method of Multipliers for Hybrid Total Variation Based Image Deblurring Model

作　　者：余瑞艳[1] YU Rui-yan(School Of Information And Mathematics, Yangtze University, Jingzhou 434020, Chin)

机构地区：[1]长江大学信息与数学学院,湖北荆州434020

出　　处：《数学的实践与认识》2016年第24期140-148,共9页Mathematics in Practice and Theory

基　　金：国家自然科学基金(6127179)

摘　　要：针对全变差模型在模糊图像复原过程中易产生振铃效应的不足,提出了图像复原的混合全变差模型.混合模型在图像边缘轮廓区域趋向为标准全变差模型,能够有效地保留边缘轮廓信息;而在光滑区域能够逼近为高阶全变差模型,达到抑制振铃效应的目的.实验结果表明,提出的混合全变差模型在复原图像结构信息的同时能够有效地抑制振铃效应的产生,得到的复原图像在客观评价标准和主观视觉效果方面均有所提高.The total variation （TV）-based deblurring model usually resulted in ringing artifacts in restored images. A hybrid total variation （HTV） model, combined by first- and second-order TVs, is proposed in this paper. In the edge regions, the HTV model could be approximated by the standard TV, and then could sufficiently preserve the edge. In contrast, this HTV model could successfully suppress the ringing artifacts in homogeneous regions. The experimental results show that the introduced model can suppress ringing artifacts while preserving the edges in the restored image. And the peak signal to noise ratio, mean structural similarity and subjective visual effect of the restored images are improved obviously.

关 键 词：图像去模糊 全变差 增广拉格朗日 交替方向乘子法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]