Stereoscopic Video Quality Assessment Based On Spatio-Temporal Features

With widespread applications of 3D technology,measuring quality for 3D multimedia plays an increasingly important role.In this paper,we propose a stereo video quality assessment framework that focuses on temporal-spatial features and stereo information.Stereoscopic video quality evaluation should take both spatial and temporal features into account,thus we start the research by spatial features analysis.We suggest that different content will change the interest area of eyes and have different assistance to distortion,thus it should be introduced into assessment system.Combined with classic image quality assessment method NIQE,we establish a basic system for video quality evaluation.In terms of temporal features,we suggest that large movement will make temporal distortion more attributive to the video quality assessment than spatial distortion.Thus we extract the motion magnitude and use it to revise spatial features and define motion consistency to judge the temporal distortion.Another important aspect of stereoscopic videos quality assessment is stereo information.By analyze pros and cons of different stereo information calculating methods,we choose binocular frame difference to evaluate the distortion of stereo information.By combining above features together,we propose a stereo video quality assessment framework that considers temporal-spatial features and stereo information.Experiment on open-source stereo videos demonstrates the prediction results of our method are highly consistent with subjective tests.