Image quality assessment based on independent component analysis
Chunheng Luo, Yang Wang, Yong Shan Ding, Zhenliang Wu · 2014
In this paper, an implementation of full reference image quality assessment based on independent component analysis (ICA) is proposed. ICA is a method for signal processing which is used here as a mathematical tool for image feature extraction. In addition to greyscale-based algorithm, we also develop an approach based on hue, saturation and value (HSV), which treats color information as an important part of the evaluation of image quality, in order to realize a more comprehensive simulation of human visual system (HVS) and thus achieve better consistency with subjective image quality assessment. Our experiment is conducted on MATLAB platform, and the results demonstrate that the HSV-based algorithm gives better performance for most distortion types such as JPEG, white noise and fast fading in comparison with some other common-used full reference image quality assessment algorithms. Besides, the influence of the size of sample windows on algorithm performance is also explored in the experiment.