基于局部离散小波变换的无参考模糊图像质量评价方法  

No-reference blur image quality assessment based on local discrete wavelet transform

作　　者：曾台英[1] 邵雪[1] 汪祖辉[1] 

机构地区：[1]上海理工大学出版印刷与艺术设计学院,上海200093

出　　处：《光电子．激光》2016年第7期754-760,共7页Journal of Optoelectronics·Laser

基　　金：国家重大科学仪器设备开发专项(2011YQ150040);新闻出版总署数字印刷工程研究中心数字传播重点实验室基金(10-00-309-000)资助项目

摘　　要：构造模糊图像的参考图像,使无参考(NR)图像质量评价转变为一种全参考(FR)图像质量评价,提出一种基于局部离散小波变换的NR模糊图像质量评价方法。首先,采用低通滤波的方法构造原始模糊图像的参考图像;然后,通过局部离散小波变换提取出原始模糊图像和参考图像的高频能量图;最后计算两幅高频能量图的相似度,得到原始模糊图像的质量。在LIVE、TID2008和CSIQ 3个数据库上的实验结果表明,本文算法的评价性能较优,相较其他常用的4个清晰度评价算法,在与主观评价的单调性和一致性上有明显提高,在LIVE数据库上的SROCC值达到0.959 8,在CISQ数据库上的SROCC值达到0.910 4,与主观评价平均意见成绩(MOS)值拟合的散点图拟合度为最优。实验数据结果表明,本文算法性能稳定,符合人类视觉系统。Construct the reference image for blur images, so that no reference image quality assessment is transformed into a full reference image quality evaluation, a no-reference blur image quality assessment based on local discrete wavelet transform is proposed. Firstly, the reference image of the original blur im- age is constructed by the method of low pass filtering;then, the high frequency energy graphs of the orig inal fuzzy image and reference image are extracted by local discrete wavelet transform;finally,the simi- larity of two high frequency energy graphs is calculated,and the quality of the original blur image is ob- tained. The experimental results on 3 databases show that the proposed algorithm is superior to the other four definition evaluation algorithms, and has a significant improvement on the monotonic and consistency with subjective evaluation. The spearman rank-order eoorelation coefficient （SROCC） value reaches 0. 959 8 in LIVE database. The performance of the with the human visual system. algorithm is stable, and the algorithm is more consistent with the human visual system.

关 键 词：无参考(NR)图像质量评价 局部离散小波变换 模糊估计 人类视觉 

分 类 号：TN919.8[电子电信—通信与信息系统]