VRL-IQA: Visual Representation Learning for Image Quality Assessment

被引:4
作者
Aslam, Muhammad Azeem [1 ]
Wei, Xu [2 ]
Ahmed, Nisar [3 ]
Saleem, Gulshan [4 ]
Amin, Tuba [5 ]
Caixue, Hui [1 ]
机构
[1] Xian Eurasia Univ, Sch Informat Engn, Xian 710065, Shaanxi, Peoples R China
[2] Chinese Acad Sci, Changchun Inst Opt Fine Mech & Phys, Changchun 130033, Jilin, Peoples R China
[3] Univ Engn & Technol, Dept Comp Engn, Lahore 54890, Pakistan
[4] Lahore Garrison Univ, Dept Comp Sci, Lahore 54000, Pakistan
[5] Govt Coll Univ Faisalabad, Dept Comp Sci, Faisalabad 38000, Pakistan
关键词
Visualization; Training; Computer architecture; Image quality; Task analysis; Representation learning; Streams; Convolutional neural networks; Transfer learning; Convolutional neural network; image quality assessment; image quality; IQA; transfer learning; visual representation learning; STRUCTURAL SIMILARITY; GRADIENT MAGNITUDE; ARCHITECTURES;
D O I
10.1109/ACCESS.2023.3340266
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
With the increasing prevalence of digital multimedia devices and the growing reliance on compression and wireless data transmission, evaluating image quality remains a persistent challenge. This study addresses the limitations of image quality assessment stemming from the expense of data annotation and the scarcity of labeled training datasets. Leveraging visual representation learning, our approach involves training a deep Convolutional Neural Network on a large image dataset generated by simulating 165 distortion scenarios across 150,000 images, resulting in 24.75 million distorted images. These distortions are labeled using an ensemble of full-reference quality assessment models. The trained model undergoes fine-tuning on diverse datasets, including TID2013, Kadid-10K, KonIQ-10K, and BIQ2021, encompassing both simulated and authentic distortions. The fine-tuning process achieves state-of-the-art image quality assessment performance, yielding Spearman's correlation coefficients of 0.921, 0.893, 0.884, and 0.793, respectively, for the four datasets. Comparative analysis with an ImageNet pre-trained model demonstrates superior performance in terms of Pearson and Spearman's correlations, achieving validation criteria with fewer epochs. These findings contribute to the advancement of IQA, offering a promising approach for robust and accurate quality prediction in various applications.
引用
收藏
页码:2458 / 2473
页数:16
相关论文
共 74 条
[1]   BIQ2021: a large-scale blind image quality assessment database [J].
Ahmed, Nisar ;
Asif, Shahzad .
JOURNAL OF ELECTRONIC IMAGING, 2022, 31 (05)
[2]   Deep ensembling for perceptual image quality assessment [J].
Ahmed, Nisar ;
Asif, H. M. Shahzad ;
Bhatti, Abdul Rauf ;
Khan, Atif .
SOFT COMPUTING, 2022, 26 (16) :7601-7622
[3]   Ensembling Convolutional Neural Networks for Perceptual Image Quality Assessment [J].
Ahmed, Nisar ;
Asif, Hafiz Muhammad Shahzad .
2019 13TH INTERNATIONAL CONFERENCE ON MATHEMATICS, ACTUARIAL SCIENCE, COMPUTER SCIENCE AND STATISTICS (MACS-13), 2019,
[4]   PIQI: perceptual image quality index based on ensemble of Gaussian process regression [J].
Ahmed, Nisar ;
Asif, Hafiz Muhammad Shahzad ;
Khalid, Hassan .
MULTIMEDIA TOOLS AND APPLICATIONS, 2021, 80 (10) :15677-15700
[5]   PERCEPTUAL QUALITY ASSESSMENT OF DIGITAL IMAGES USING DEEP FEATURES [J].
Ahmed, Nisar ;
Asif, Hafiz Muhammad Shahzad .
COMPUTING AND INFORMATICS, 2020, 39 (03) :385-409
[6]  
[Anonymous], 2002, Tech. rep
[7]  
Chaofeng Li, 2009, Proceedings of the SPIE - The International Society for Optical Engineering, V7242, DOI 10.1117/12.811821
[8]   No-reference color image quality assessment: from entropy to perceptual quality [J].
Chen, Xiaoqiao ;
Zhang, Qingyi ;
Lin, Manhui ;
Yang, Guangyi ;
He, Chu .
EURASIP JOURNAL ON IMAGE AND VIDEO PROCESSING, 2019, 2019 (01)
[9]  
Cole E., 2022, PROC IEEECVF C COMPU, P01
[10]  
Deng J, 2009, PROC CVPR IEEE, P248, DOI 10.1109/CVPRW.2009.5206848