A Visual Perceptual Bayesian Theory for Stereoscopic Images’ Quality Assessment

Abstract

Stereoscopic image quality measurement (SIQM) is a fundamental and challenging issue in image processing. In this letter, we present a full-reference (FR) SIQM method based on the visual perceptual Bayesian theory which consists of binocular summation and difference channels. Specifically, we first apply the contrast sensitivity filtering to each image of both the reference and distorted stereo pairs. Constructively, a new cyclopean image is generated by considering the binocular perceptual model and binocular rivalry simultaneously. Afterward, the qualities of the summation image, difference image, and cyclopean image between reference and distorted stereo pairs are computed by using structural similarity index to form the underlying quality-ware features. Finally, the kernel ridge regression is used to simulate a nonlinear relationship between the quality-aware features and objective quality scores. Experimental results demonstrate that the proposed method achieves high consistency with human opinions and outperforms several state-of-the-art FR-SIQM methods.