基于空-频域印刷混合失真无参考质量评价  

No-reference Quality Assessment Method of Printing Multiply Distort Image Based on Spatial Frequency Domain

作　　者：麻祥才 肖颖[1] 钱志伟[1] 邹来涛 王晓红[2] 韩双红 卢辉 丁桂芝 MA Xiang-cai;XIAO Ying;QIAN Zhi-wei;ZOU Lai-tao;WANG Xiao-hong;HAN Shuang-hong;LU Hui;DING Gui-zhi(Shanghai Publishing and Printing College, Shanghai 200093, China;University of Shanghai for Science andTechnology, Shanghai 200093, China;Shanghai Printing Trade Association, Shanghai 200093, China)

机构地区：[1]上海出版印刷高等专科学校,上海200093 [2]上海理工大学,上海200093 [3]上海市印刷行业协会,上海200093

出　　处：《包装工程》2019年第9期224-230,共7页Packaging Engineering

基　　金：上海市教育发展基金会和上海市教育委员会"晨光计划"(18CGB09);"柔版印刷绿色制版与标准化实验室"招标课题(ZBKT201809)

摘　　要：目的为了解决数字印刷模糊混合失真图像质量评价算法与人眼视觉主观评价一致性较低的问题。方法基于SSEQ算法思想在空间域和频率域中提取混合失真图像的信息熵、亮度方差、频域熵以及高频系数奇异值改变量关键特征,利用支持向量机训练构建印刷模糊混合失真图像的无参考评价模型。结果客观验证了25幅图像,算法客观计算得分与主观评价得分的SROCC系数为0.94,PLCC系数为0.93,主客观一致性较高。结论所提算法与人眼视觉感知具有很好的主观一致性,可以用于图像等印刷品的质量客观评价。The work aims to solve the problem that the blurred image of digital printing multiply distorted quality assessment algorithm has low consistency with the subjective evaluation of human perception. Key characteristics including entropy, brightness variance, frequency entropy and the high frequency coefficient singular value change were extracted from distorted image in spatial and frequency domain based on the idea of SSEQ algorithm to establish the no-reference image quality assessment method with support vector machine in this paper. 25 images were verified objectively and the SROCC coefficient and PLCC coefficient of the objective calculation score and the subjective evaluation score of the algorithm were 0.94 and 0.93 respectively, with high subjective and objective consistency. The algorithm is in good subjective consistency with the ocular perception and suited for the printed image quality assessment.

关 键 词：数字印刷图像 混合失真 无参考图像质量评价 

分 类 号：TS807[轻工技术与工程]