A Perceptual Video Quality Model Based on Visual Attention Variation

This article establishes the video quality assessment model based on variation in visual attention, from coding features and video content characteristics of the two aspects. We explore the impact of the quantization parameter in the respect of coding and surprising parameters to measure the attraction for human in the respect of human visual system according to the extent of the surprising events in different content video, then establish the eventual quality assessment model related to the quantization step and surprising parameter based on content. Experimental results show that the Pearson Correlation coefficient between objective score obtained from the model calculations and subjective MOS scores is as high as 0.9740 and the Root Mean Square Error of them is 0.1575. The proposed model has a great accuracy on predicting users' QoE(quality of experience) of videos of different content account with its good performance.