An objective visual comfort prediction metric of stereoscopic images based on stereoscopic saliency model

摘要

In this paper, a simple but effective objective visual comfort prediction metric for stereoscopic images is presented. In this metric, stereoscopic saliency map is first calculated from the stereoscopic images and the corresponding disparity maps using region covariance. Then, visual discomfort perceptual features are obtained by using the stereoscopic saliency map as weighting. Finally, support vector regression is performed to predict the visual comfort score by establishing the function between perceptual visual comfort features and mean opinion scores. Experimental results on the stereoscopic image database demonstrate that the proposed metric outperforms other two traditional objective methods in predicting visual comfort with regard to the commonly used statistical criteria, namely PLCC, SROCC, KROCC, and RMSE.