电子学报 ›› 2016, Vol. 44 ›› Issue (4): 804-812.DOI: 10.3969/j.issn.0372-2112.2016.04.008

• 学术论文 • 上一篇    下一篇

结合结构信息和亮度统计的无参考图像质量评价

沈军民1, 李俊峰2, 戴文战3   

  1. 1. 浙江理工大学电子信息工程系, 浙江杭州 310018;
    2. 浙江理工大学自动化系, 浙江杭州 310018;
    3. 浙江工商大学电子信息工程系, 浙江杭州 310018
  • 收稿日期:2014-08-08 修回日期:2015-07-22 出版日期:2016-04-25
    • 通讯作者:
    • 李俊峰
    • 作者简介:
    • 沈军民 男,1976年6月出生,浙江金华人.2003年和2011年分别在中国计量科学研究院和东华大学获工学硕士和工学博士学位.现为浙江理工大学讲师,主要从事图像质量评价、机器视觉及产品检测等方面的研究工作. E-mail:sjumi1976@163.com;戴文战 男,1958年8月出生,浙江台州人.1982年、1988年分别在浙江大学、华东理工大学获工学学士、工学硕士学位.现为浙江工商大学教授、博士生导师,主要从事视频图像处理、图像融合与系统工程. E-mail:dwz@zjgsu.edu.cn
    • 基金资助:
    • 国家自然科学基金 (No.61374022); 浙江省公益性技术应用研究计划 (No.2014C33109)

No-Reference Image Quality Assessment Based on Structure Information and Luminance Statistics

SHEN Jun-min1, LI Jun-feng2, DAI Wen-zhan3   

  1. 1. Department of Electronic Information Engineering, Zhejiang Sci-Tech University, Hangzhou, Zhejiang 310018, China;
    2. Department of Automation, Zhejiang Sci-Tech University, Hangzhou, Zhejiang 310018, China;
    3. Department of Electronic Information Engineering, Zhejiang Gongshang University, Hangzhou, Zhejiang 310018, China
  • Received:2014-08-08 Revised:2015-07-22 Online:2016-04-25 Published:2016-04-25

摘要:

基于非下采样Contourlet变换(Nonsubsampled Contourlet Transform,NSCT)子带系数间的结构相关性,本文提出了一种通用的无参考图像质量评价方法.首先,利用互信息分析NSCT子带系数间的相关性,确定出相关性比较强的子带系数;其次,分别计算这些子带系数间的结构信息比较算子,以此作为描述图像结构相关性的统计特征;进而,结合空间域亮度均值减损对比归一化(Mean Subtracted Contrast Normalized,MSCN)系数及其邻域系数的统计特征,分别构造相应的无参考图像质量评价模型和图像失真类型识别模型;最后,在LIVE等图像质量评价数据库上进行了大量的实验仿真.仿真结果表明,评价模型的评价结果与人类主观评价具有非常高的相关性,与当今主流评价算法相比非常具有竞争性.

关键词: 无参考图像质量评价, 非下采样Contourlet变换, 结构相似度, 亮度统计

Abstract:

Based on the correlation between subband coefficients of nonsubsampled contourlet transform(NSCT),a general-purpose no-reference image quality assessment(NR-IQA)method is proposed.Firstly,the correlation between NSCT subband coefficients was analyzed using mutual information and the subband coefficients with strong correlation were determined.Secondly,the structure comparison operator of those subband coefficients were calculated respectively and used to describe the statistics characteristic of image structure information.Moreover,a no-reference image quality assessment model and an image distortion type recognition model were constructed by combining the statistical features of the mean subtracted contrast normalized(MSCN)coefficients and the structural similarity of NSCT coefficients.Finally,a large number of simulation experiments were carried out in the LIVE image quality evaluation database.The simulation results show that this method is suitable for many common image distortion types and correlates well with the human judgments of image quality,and the assessment model is competitive with the nowadays' mainstream evaluation method.

Key words: no-reference image quality assessment, nonsubsampled contourlet transform, structural similarity, luminance statistics

中图分类号: