全参考图像质量评价回顾与展望  

Review and Prospect of Full Reference Image Quality Assessment

作　　者：高敏娟 党宏社[1] 魏立力 刘国军 张选德 GAO Min-juan;DANG Hong-she;WEI Li-li;LIU Guo-jun;ZHANG Xuan-de(School of Electrical and Control Engineering,Shaanxi University of Science and Technology,Xi’an,Shaanxi 710021,China;School of Mathematics and Statistics,Ningxia University,Yinchuan,Ningxia 750021,China;School of Electronic Information and Artificial Intelligence,Shaanxi University of Science and Technology,Xi’an,Shaanxi 710021,China)

机构地区：[1]陕西科技大学电气与控制工程学院,陕西西安710021 [2]宁夏大学数学统计学院,宁夏银川750021 [3]陕西科技大学电子信息与人工智能学院,陕西西安710021

出　　处：《电子学报》2021年第11期2261-2272,共12页Acta Electronica Sinica

基　　金：国家自然科学基金(No.61871260);陕西省自然科学基金(No.2020JM-509)。

摘　　要：全参考图像质量评价(Full Reference Image Quality Assessment,FR-IQA)是IQA领域广为研究的类型之一.本文回顾了FR-IQA的发展历程,对FR-IQA应用现状和通用FR-IQA问题的构建进行综述,以及对FR-IQA算法进行总结和梳理.并在此基础上,重点分析了现有研究中存在的问题,包括问题构建的合理性、建模的全面性问题、知识驱动与数据驱动结合的问题等.基于对主观评价过程的深入分析,结合现有研究存在的问题,探讨了主观评分采用模糊建模和知识与数据联合驱动构建算法两个可能的研究方向,以期对后续的研究者提供参考.Full reference image quality assessment(FR-IQA)is one of the types widely studied in the field of IQA.This paper reviews the development of FR-IQA,summarizes the application status of FR-IQA and the construction of gener⁃al FR-IQA problems,and summarizes and combs FR-IQA algorithms.And on this basis,it focuses on the problems existing in the existing research,including the rationality of the problem construction,the comprehensive problem of modeling,and the problem of the combination of knowledge-driven and data-driven.Finally,based on the in-depth analysis of the subjec⁃tive evaluation process and the existing problems in the existing research,the two possible research directions of subjective scoring using fuzzy modeling and knowledge-data-driven construction algorithm are discussed,in order to provide refer⁃ence for subsequent researchers.

关 键 词：图像质量评价 FR-IQA 主观评分 模糊建模 知识与数据联合驱动 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]