No-reference image quality assessment based on gradient histogram response. (August 2016)

• Record Type: Journal Article
• Authors: Sun, Tongfeng; Ding, Shifei; Chen, Wei; Xu, Xinzheng
• Is Part Of: Computers & electrical engineering. Volume 54(2016)
• Page Start: 330
• Page End: 344
• Publication Date: 2016-08
• Subjects: Gradient histogram response -- Object input -- Object output -- Local image transform -- No-reference image quality assessment; Computer engineering -- Periodicals; Electrical engineering -- Periodicals; Electrical engineering -- Data processing -- Periodicals; Ordinateurs -- Conception et construction -- Périodiques; Électrotechnique -- Périodiques; Électrotechnique -- Informatique -- Périodiques; Computer engineering; Electrical engineering; Electrical engineering -- Data processing; Periodicals; Electronic journals; 621.302854
• DOI: 10.1016/j.compeleceng.2015.11.007
• Languages: English
• ISSNs: 0045-7906

Abstract

Highlights: A NR image quality assessment based on gradient histogram response (GHR) is proposed. A test image is preprocessed to produce a noise image object and a blur image object. GHR is the gradient histogram variation of an image object under a local transform. GHR performs well for the images degraded by different distortions or mixed distortions. Graphical abstract: Abstract: In view of the fact that objects with different natures usually respond differently to the same external stimulus, this paper proposes a no-reference image quality assessment based on gradient histogram response (GHR). GHR is the gradient histogram variation of an image object under a local transform. In the metric, through preprocessing, a test image is transformed to a noise image and a blur image, which are taken as two image objects. Each image object is exerted with a local transform as an object input, and its GHR as an object output is extracted in multiscale space. The two GHRs compose a global feature vector and are mapped to an image quality score. Experiments show that GHR outperforms state-of-the-art no-reference metrics statistically in the condition that test images are degraded by different types of distortions. Especially, the metric is feasible for the quality assessment of the images degraded by mixed distortions though the types of these images are not included in the training database.