应用再模糊理论的无参考图像质量评价算法  

Evaluation algorithm of non-reference image qualityapplying re-fuzzy theory

作　　者：俞胜 张怀亮[1,2] 彭玲[2] 罗永康 YU Sheng;ZHANG Huailiang;PENG Ling;LUO Yongkang(State Key Laboratory of High Performance and Complex Manufacturing,Central South University,Changsha 410083,China;College of Mechanical and Electrical Engineering,Central South University,Changsha 410083,China)

机构地区：[1]中南大学高性能复杂制造国家重点实验室,湖南长沙410083 [2]中南大学机电工程学院,湖南长沙410083

出　　处：《传感器与微系统》2021年第8期140-143,共4页Transducer and Microsystem Technologies

基　　金：国家重点基础研究发展计划(973计划)资助项目(2013CB035400)。

摘　　要：为了提高无参考图像质量评价方法的评价性能,提出了一种应用图像再模糊理论的图像质量评价算法。将待测图像经过再模糊化处理,利用模糊前后两张图像的差异提取图像的局部特征和全局特征,共同构建多维特征向量,并采用支持向量回归(SVR)的方法实现图像无参考质量评价。在公开的图像数据库中进行测试实验,结果表明:所提图像质量评价方法具有较高的准确性和较好的泛化能力,与人的主观评价具有较好的一致性。To improve evaluation performance of the non-reference image quality evaluation method,an image quality evaluation algorithm applying image re-fuzzy theory is proposed.The image to be tested is subjected for re-fuzzy process,extract the local and global features of the image by using difference between the two images before and after blurring,co-building multi-dimensional feature vectors,and the support vector regression(SVR)method is used to realize the non-reference quality evaluation of image.Test experiments in public image database show that the image quality evaluation method has higher accuracy and better generalization ability,and has better consistency with human subjective evaluation.

关 键 词：再模糊理论 无参考图像质量评价 支持向量回归 局部特征 全局特征 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]