Title :
Blind Image Quality Assessment Using Natural Scene Statistics in the Gradient Domain
Author :
Tonghan Wang ; Huazhong Shu ; Huizhen Jia ; Baosheng Li ; Lu Zhang
Author_Institution :
Lab. of Image Sci. & Technol., Southeast Univ., Nanjing, China
Abstract :
An efficient, general-purpose, blind/no-reference image quality assessment (NR-IQA) algorithm based on natural image statistics in the gradient domain is proposed in this letter. We call it REFIINGS (REFerrenceless Image Integrity Notator using Gradient Statistics). The gradient of an image describes its geometric features which can be easily captured by the human visual system (HVS). In the literature, gradient-relevant methods have gotten big success in full-reference (FR) IQA and reduced-reference (RR) IQA. Inspired by these, we extend it to NR-IQA. REFIINGS utilizes the parameters of generalized Laplace distribution as part of its features, and the parameters are directly computed using given formulas which avoid parameters estimation. REFIINGS is computationally quite efficient which makes it an attractive option for the use in real-time blind assessment of visual quality. When tested on the benchmark image database, our method is quite promising.
Keywords :
feature extraction; gradient methods; statistics; visual databases; FR IQA; HVS; NR-IQA algorithm; REFIINGS; RR IQA; benchmark image database; blind image quality assessment; full-reference IQA; generalized Laplace distribution; geometric features; gradient domain; human visual system; natural image statistics; natural scene statistics; reduced-reference IQA; referenceless image integrity notator using gradient statistics; visual quality real-time blind assessment; Entropy; Feature extraction; Histograms; Image quality; Nonlinear distortion; Transform coding; Visualization; Image quality assessment; generalized Laplace distribution; gradient domain; natural scene statistics; no reference;
Conference_Titel :
Modelling Symposium (AMS), 2014 8th Asia
Print_ISBN :
978-1-4799-6486-4
DOI :
10.1109/AMS.2014.22