Title:	A FEATURE SIMILARITY INDEX BASED ON THE OPPONENT WEIGHTING FUNCTION FOR IMAGE QUALITY ASSESSMENT
Author(s):	Chengho Hsin, Zheng Chiu Chen and Shaw-Jyh Shin
ISBN:	978-989-8533-66-1
Editors:	Yingcai Xiao and Ajith P. Abraham
Year:	2017
Edition:	Single
Keywords:	Image quality assessment, gradient similarity, weighted pooling, full reference
Type:	Full Paper
First Page:	131
Last Page:	138
Language:	English
Cover:
Full Contents:	click to dowload
Paper Abstract:	Automatic image quality assessment (IQA) plays a vital role in various image and video processing applications. The most successful approach is based on the concept of structural similarity. In this paper, a high-performance full reference IQA model based on this approach is proposed that utilizes joint feature similarity and opponent weighted pooling. Three types of features including gradient magnitude, high-pass filtered component, and luminance mean are used altogether to establish a joint similarity map, which can account for most of the distortions encountered in real applications. By considering the shortcomings of the commonly used weighted pooling, a novel opponent weighting function is devised to assign larger weights to the distorted structures, smaller weights to those distortion-free structures. The experimental results on three image databases show that the proposed index provides comparable or better predictions than the competing state-of-the-art IQA metrics in the literature, it is effective and reliable.
	Go Back