基于变分的盲图像复原质量评价指标  

A Modern Image Quality Measurement Method for Blind Image Restoration

作　　者：成孝刚[1] 安明伟[1] 阮雅端[1] 陈启美[1] 

机构地区：[1]南京大学电子科学与工程学院,南京210093

出　　处：《自动化学报》2013年第4期418-423,共6页Acta Automatica Sinica

基　　金：国家科技重大专项(2012ZX03005012;2009ZX03003-007;2011ZX03005-004-03)资助~~

摘　　要：盲图像复原过程中,图像质量评价至关重要.通过分析重构图像质量与其总变分值之间的关系,提出了用于图像复原的一种基于总变分(Total bounded variation,TBV)的图像质量评估方法,并构建关系模型,证明了原始清晰图像的总变分值在所有模糊图像中具有极大值,且在所有重构图像的变分值中具有极小值.通过分析,得出结论:当总变分取极值时,基于所提度量方法,可以获得更好的盲图像重构效果.最后,比较了原始清晰图像、模糊图像和重构图像之间的变分值,计算机仿真验证了该方法的有效性和准确性.In the process of blind image restoration, image quality assessment is of paramount importance. In this paper, A novel image quality assessment method is presented by analyzing the relation between reconstructed image quality and its total bounded variation （TBV）, on this basis, the relationship model is constructed, that is, the original clear image＇s TBV is maximum in all the blurring image, and it is minimal in all the reconstructed image. Further, based on the metric method proposed, the better blind image reconstruction effect is obtained when the TBV is extremal. Finally, the TBV of original clear image, blurred images and blind restored images are compared, the simulation results shows the validation and veracity of the method proposed.

关 键 词：总变分 图像质量 盲图像复原 评价函数 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]