Perceptual Quality Assessment For Asymmetrically Distorted Stereoscopic Video

With the rapid development of digital devices and the increasing diversification of media,the stereoscopic video has gradually become an essential form of multimedia application.The data volume of the stereoscopic video is much higher than that of the traditional 2D video.Balancing the massive amount of stereoscopic video data with limited network bandwidth and providing a better quality of experience is crucial for advancing the stereoscopic video industrialization.Therefore,research on stereoscopic video quality assessment is important.However,it is difficult to understand how the human visual system(HVS)perceiving the stereoscopic video.The visual quality of the stereoscopic video is not only depended on the fidelity of the video content,but also related to the human visual mechanisms,such as binocular rivalry.Specially,the asymmetrical distortions between the left-and right-views of the stereoscopic video are easy to cause binocular rivalry,which might result in a deviation in the perceived quality.Therefore,the perceptual quality assessment of asymmetrically distorted stereoscopic video is challenging.Existing stereoscopic image/video quality assessment methods are not sufficient to provide reasonably accurate quality predictions of stereoscopic videos.Besides,there is still a lack of research on perceptual quality assessment of asymmetrically distorted stereoscopic video,which inspired this paper to carry out this study.The specific research content is as follows:(1)Asymmetrically distorted stereoscopic video quality assessment using a twostage quality weighting method.By considering that existing asymmetrically distorted stereoscopic video quality assessment models do not consider temporal information,this paper proposes an asymmetrically distorted stereoscopic video quality assessment model based on a two-stage quality weighting,which considers both the spatial and temporal information of the stereoscopic video.The first-stage weighting aims to extract effective frame-level spatial and temporal features,which are used as the weighting factors to fuse the spatial and temporal distortions for the final visual quality of the single-view video.To simulate the perception difference of the HVS to different visual stimuli,the second-stage weighting method aims to extract effective video-level spatial and temporal features to establish a temporal binocular rivalry inspired model,which can adaptively quantify the weighting of the single-view video in binocular rivalry.The final visual quality score of the stereoscopic video is computed by fusing the quality scores of the left-and right-view videos.(2)Asymmetrically distorted stereoscopic video quality assessment by considering the motion variation.To extend the method(1)to be more robust for the quality assessment of stereoscopic video with complex scenes,such as diverse motion,this paper proposes a novel method for the asymmetrically distorted stereoscopic video quality assessment by considering the motion variation.Visual perception studies have shown that the moving objects is easy to attract visual attention,which leads to low sensitivity of the HVS to the background.To simulate such HVS characteristic,this paper first determines the motion type of the consecutive frames of stereoscopic video through the coefficient of variation(CV),and then designs the methods of quantifying perceptual visual information content.Finally,the perceptual visual information content of the single-view video is used as the weighting factor to combine the quality scores of the left-and right-view videos for computing the final visual quality score of the stereoscopic video.