Blind image quality assessment based on statistics features and perceptual features

Article type: Research Article

Authors: Zhao, Youen^{a; b; *} | Ji, Xiuhua^{a; b} | Liu, Zhaoguang^{a; b}

Affiliations: [a] School of Computer Science and Technology, Shandong University of Finance and Economics, Jinan, China | [b] Provincial Key Laboratory of Digital Media Technology, Shandong University of Finance and Economics, Jinan, China

Correspondence: [*] Corresponding author. Youen Zhao, E-mail: [email protected].

Abstract: Blind image quality assessment (BIQA) aims to evaluate the quality of an image without information regarding its reference image. In this paper, we proposed a novel BIQA method, which combines thirty six natural scene statistics (NSS) features, two color statistics features and four perceptual features to construct an image quality assessment model. Support Vector Regression (SVR) is adopted to build the relationship between these features and image quality scores, yielding a measure of image quality. Experimental results in LIVE, TID2013 databases and their cross validations show that the proposed method records a higher correlations with human subjective judgments of visual quality and delivers highly competitive performance with state-of-the-art BIQA models.

Keywords: Blind image quality assessment, natural scene statistics feature, perceptual feature, color statistics feature, support vector regression

DOI: 10.3233/JIFS-190998

Journal: Journal of Intelligent & Fuzzy Systems, vol. 38, no. 3, pp. 3515-3526, 2020