New Reduced-Reference Stereo Image Quality Assessment Model for 3D Visual Communication.

Visual quality assessment of stereo image plays an important role in three dimensional visual communication. Considering the processing of binocular perception in viewing stereo image, we present a reduced-reference (RR) stereo image quality assessment (SIQA) model based on binocular perceptual characteristics. Firstly, stereo images are divided into binocular fusion portion and binocular rivalry portion with internal generative mechanism. Then, cyclopean view is generated according to the binocular fusion portion and binocular rivalry portion with binocular perception, and the Gaussian scale mixture RR features from cyclopean view and the binocular rivalry portion for SIQA. Finally, the quality indicators of cyclopean view and binocular rivalry portion are computed to obtain the final SIQA score. The proposed model is tested on the LIVE 3D IQA database. Experimental results show that compared with the state-of-the-art methods, the proposed model has high correlation with subjective perception and can evaluate human stereo visual properties effectively.