Binocular visual characteristics-based stereoscopic image quality assessment metric for 3D video system

Abstract

Compared with two-dimensional image quality assessment 2D-IQA, stereoscopic IQA SIQA has to take the binocular perception into consideration. In this paper, a new SIQA metric is proposed for three-dimensional video systems, which is based on binocular visual characteristics and contains two parts: binocular perceptual distortion evaluation BPDE and stereoscopic evaluation SE. To get the results of the BPDE and SE, a simple-complex cell along with different human visual system characteristics are used to compute the quality of binocular fusion maps and disparity maps between original and distorted stereoscopic image pairs. The final quality score is obtained by a nonlinear combination of BPDE value and SE value. Experimental results show that proposed SIQA metric in which the Pearson linear correlation coefficient PLCC and Spearman rank order correlation coefficient SROCC exceed 0.94 and 0.93 outperform current 2D-IQA and SIQA metrics.