[1]高方远,何立火.基于深度网络和视觉特性的无参考图像质量评价方法[J].南京师范大学学报(自然科学版),2019,42(03):20-26.[doi:10.3969/j.issn.1001-4616.2019.03.003]
　Gao Fangyuan,He Lihuo.No-reference Image Quality Assessment Based on DeepNetwork and Visual Characteristics[J].Journal of Nanjing Normal University(Natural Science Edition),2019,42(03):20-26.[doi:10.3969/j.issn.1001-4616.2019.03.003]
点击复制

基于深度网络和视觉特性的无参考图像质量评价方法()
分享到：

《南京师范大学学报》（自然科学版）[ISSN:1001-4616/CN:32-1239/N]

卷:
第42卷
期数:
2019年03期
页码:
20-26
栏目:
·全国机器学习会议论文专栏·
出版日期:
2019-09-30

文章信息/Info

Title:
No-reference Image Quality Assessment Based on DeepNetwork and Visual Characteristics
文章编号:
1001-4616(2019)03-0020-07
作者:
高方远1; 何立火2
(1.北京航空航天大学数学与系统科学学院,北京 102206)(2.西安电子科技大学电子工程学院,陕西 西安 710071)
Author(s):
Gao Fangyuan1; He Lihuo2
1.School of Mathematics and Systems Science,Beihang University,Beijing 102206,China)(2.School of Electronic Engeering,Xidian University,Xi’an 710071,China
关键词:
图像质量评价; 深度网络; 视觉特性; 最差情况加权策略
Keywords:
image quality assessment; deep network; visual characteristic; worst-case pooling mechanism
分类号:
TP391
DOI:
10.3969/j.issn.1001-4616.2019.03.003
文献标志码:
A
摘要:
图像质量评价是图像处理和计算机视觉领域的基础性问题,对于视觉信息的采集、处理和分析系统性能的评判具有重要的意义. 现有的无参考型图像质量评价方法都是基于自然统计规律的,或者构建单一网络模型,并未考虑视觉感知特性,使得最终的评价结果与主观感受间存在较大差异. 为此,本文提出一种结合多种网络特性和最差视觉感知特性的无参考型图像质量评价方法. 首先,提取图像的去均值对比度归一化特征,将特征图进行重叠分块; 然后,构建VGG与Inception相结合的深度网络,对图像块提取深度感知特征; 最后,将分块图像的质量分数集合进行排序,利用视觉感知最差情况加权策略对序列进行加权求和,得到最终的图像质量分数. 在国际公开的质量评价数据库CSIQ、LIVE和TID2013上的实验结果表明,本文方法取得了优于现有方法的主客观一致性性能.
Abstract:
Image quality assessment(IQA)is a fundamental problem in image processing and computer vision. IQA is very important for the performance evaluation of acquisition,processing and analysis systems of visual information. Most of the existing no-reference(NR)image quality assessments are based on the natural statistical characteristics,or design deep network to predict the image quality. These methods have not considered the visual characteristics,which produces the large difference in image quality score between the objective and subjective methods. To overcome this problem,this paper proposes a NR-IQA method based on the deep network and visual characteristics. Firstly,mean subtracted contrast normalized map of the image is calculated,and the map is resampled into overlapped patches randomly. Then the deep network is designed by combining the VGG net and Inception net to predict the quality scores of patches. Finally,the scores are ranked and the worst-case pooling is utilized to weight the scores to obtain the image quality score. Experimental results on public CSIQ,LIVE and TID2013 databases show that the proposed method performs consistently with the subjective perception and has a better performance than state-of-the-art methods.

参考文献/References:

[1] WANG Z,BOVIK A C. Modern image quality assessment[M]. San Rafael,CA,USA:Morgan & Claypool,2006.
[2]WANG Z,BOVIK A C,SHEIKH H R,et al. Image quality assessment:from error visibility to structural similarity[J]. IEEE transactions on image processing,2004,13(4):600-612.
[3]ZHANG D,ZHANG L,MOU X Q,et al. FSIM:a feature similarity index for image quality assessment[J]. IEEE transactions on image processing,2011,20(8):2378-2386.
[4]SHEIKH H R,BOVIK A C. Image information and visual quality[J]. IEEE transactions on image processing,2006,15(2):430-444.
[5]HE L H,TAO D G,LI X L,et al. Sparse representation for blind image quality assessment[C]//IEEE Conference on Computer Vision and Pattern Recognition(CVPR). Providence,Rhode Island,USA,2012:1146-1153.
[6]SHEIKH H R,BOVIK A C,CORMACK L. No-reference quality assessment using natural scene statistics:JPEG2000[J]. IEEE transactions on image processing,2005,14(11):1918-1927.
[7]KUNDU D,EVANS B L. Spatial domain synthetic scene statistics[C]//48th Asilomar Conference on Signals,Systems and Computers(Asilomar). Pacific Grove,California,USA,2014:948-954.
[8]TANG H X,NEEL J,ASHISH K. Blind image quality assessment using semi-supervised rectifier networks[C]//IEEE Conference on Computer Vision and Pattern Recognition(CVPR). Columbus,Ohio,USA,2014:2877-2884.
[9]KANG L,YE P,LI Y,et al. Convolutional neural networks for no-reference image quality assessment[C]//IEEE Conference on Computer Vision and Pattern Recognition(CVPR). Columbus,Ohio,USA,2014:1733-1740.
[10]KANG L,YE P,LI Y,et al. Simultaneous estimation of image quality and distortion via multi-task convolutional neural networks[C]//IEEE International Conference on Image Processing(ICIP). Quebec City,Canada,2015:2791-2795.
[11]KIM J,LEE S. Fully deep blind image quality predictor[J]. IEEE journal of selected topics in signal processing,2017,11(1):206-230.
[12]REN H Y,CHEN D Q,WANG Y Z. RAN4IQA:restorative adversarial nets for no-reference image quality assessment[C]//The Thirty-Second AAAI Conference on Artificial Intelligence(AAAI). New Orleans,Louisiana,USA,2018:7308-7314.
[13]BOSSE S,MANIRY D,MüLLER K,et al. Deep neural networks for no-reference and full-reference image quality assessment[J]. IEEE transactions on image processing,2018,27(1):206-219.
[14]KIM J,NGUYEN A,LEE S. Deep CNN-based blind image quality predictor[J]. IEEE transactions on neural networks and learning systems,2019,30(1):11-24.
[15]SHEIKH H R,WANG Z,CORMACK L,et al. Live image quality assessment database release 2(LIVE II)[DB/OL]. [2019-01-18]. http://live.ece.utexas.edu/research/quality.
[16]LARSON E C,CHANDLER D M. Categorical image quality(CSIQ)database[DB/OL]. [2019-01-18]. http://vision.okstate.edu/csiq.
[17]PONOMARENKO N,IEREMEIEV O,LUKIN V,et al. A new color image database TID2013:Innovations and Results[M]//Blanc-Talon J,Kasinski A,Philips W,et al. Advanced Concepts for Intelligent Vision Systems. Berlin,Heidelberg:Springer-Verlag,2013.
[18]VQEG. Validation of reduced-reference and no-reference objective models for standard definition television,Phase I[DB/OL]. [2019-01-18]. http://www.vqeg.org/.

相似文献/References:

[1]刘大瑾,叶建兵,刘家骏.SSIM框架下基于SVD的灰度图像质量评价算法研究[J].南京师范大学学报(自然科学版),2017,40(01):73.[doi:10.3969/j.issn.1001-4616.2017.01.011]
　Liu Dajin,Ye Jianbing,Liu Jiajun.SVD-Based Gray-Scale Image Quality AssessmentAlgorithms in the SSIM Perspective[J].Journal of Nanjing Normal University(Natural Science Edition),2017,40(03):73.[doi:10.3969/j.issn.1001-4616.2017.01.011]

备注/Memo

备注/Memo:
收稿日期:2019-07-05.基金项目:国家自然科学基金(61876146). 通讯联系人:何立火,博士,副教授,研究方向:图像质量评价. E-mail:lhhe@mail.xidian.edu.cn
更新日期/Last Update: 2019-09-30