摘要
随着数字图像技术的发展图像质量评价已成为了一个重要的研究课题。作为图像质量的一部分,图像逼真度通常由均方误差来表示。均方误差作为既简单又古老的客观图像质量评价方法沿用至今,在图像质量评价中扮演着重要角色,但在实际应用过程中发现其存在一定的局限性。从对比度拉伸、亮度变化、加噪声、模糊、放缩、平移、旋转等经典图像变形的角度出发,实现了给定一个均方误差值和允许误差值,得到11种均方误差值近似相等的变形图像。经分析得出,同一均方误差下的不同变形图像给人的主观感受不同;图像经很小的几何修正后,对应的均方误差值可能很大;均方误差越大,图像失真越严重。并设计了一个120人的主观评价实验,验证了均方误差在基础变形图像上评价的局限性。
With the development of digital image technology, image quality evaluation has become an important research subject. As a part of image quality, image fidelity is usually represented by mean square error（ MSE）.MSE, as a simple and ancient objective image quality evaluation method, plays an important role in image quality evaluation. But it has some limitations in practical application. From the angle of contrast stretching, brightness change, additive noise, fuzzy, scaling, translation, rotation and other classic image deformation, 11 kinds of deformation images were achieved whose MSE was approximately equal by given a MSE and allowable error value. The analysis shows that the different deformation images give different subjective feelings under the same MSE. After a very small geometric correction of the image, the change of corresponding MSE may be very large. The greater the MSE is, the more serious the image distortion is. A 120 person＇s subjective evaluation experiment is designed, which verifies the limitations of the MSE in the basic deformation images.
作者
郑艳梅
岳向江
李翠
芦碧波
ZHENG Yan-mei;YUE Xiang-jiang;LI Cui;LU Bi-bo(College of Computer Science and Technology-,Henan Polytechnic University,Jiaozuo 454000,China;,School of Software and Mieroeleetronie,Northwestern Polyteehnieal University,Xi＇an 710072,China)
出处
《测控技术》
CSCD
2018年第8期117-120,共4页
Measurement & Control Technology
基金
国家自然科学基金(11301113
51174263)
河南省教育厅科技攻关重点项目(14A520029
16A520053)
河南理工大学博士基金资助项目(B2016-40)
关键词
均方误差
质量评估
变形图像
主观评价
mean square error（MSE）
quality evaluation
deformation image
subjective evaluation
