| With the commercial application of LTE, the Age of Big Data is approaching and will bring great challenges as well as opportunities to the mobile communication industry.In the future, video service will be one of the main services of mobile communication. However, the transmission of digital video may produce a variety of distortions in the received video and lead to damages on video quality.In order to ensure the Quality of Service (QoS) of video services, a complete and effective video quality assessment system should be set up to evaluate video QoE accurately and quickly. This paper mainly does some researchon Human Visual System (HVS) based video QoE evaluation related technology in wireless network.Firstly, this paper analyzes spatial and temporal traits of video sequences in wireless network, models the evaluation scheme for video features and summarizes the principle of visual attention mechanism associated with the HVS. Next, this paper deeply studies the visual information distribution in video sequences, proposes an extraction model through signal transformation and traits fusion, presents and analyzes the visual saliency mapping model.Thirdly, we study the application of Temporal-Frequency Transform (TFT) in image processing and present an image visual information extraction model based on TFT. At last, the visual information based video QoE estimation model is proposed after the research on visual attention mechanism, including a spatial model for the image quality pooling and a temporal model for the frame quality summation.The visual attention mechanism mentioned herein is based on the response of HVS to the video traits stimuli. At the beginning, this paper models and analyzes the evaluation scheme of video sequences features, in which video sequences traits are measured by visual information. Then a Perceptual Stimuli-Response Model (PSRM) is proposed to map visual information to visual saliency, which is taken as a direct way to illustrate the distribution of visual attention. The experimental results show that the proposed visual attention extraction scheme associated with visual information can extract Region of Interest (ROI) effectively and reduce computational complexity.The proposed video QoE metric consists of two steps:spatial and temporal pooling. The former one is to calculate a frame quality by pooling the frame’s image quality according to the visual attention distribution. The latter one is to obtain the overall video quality through summation along the temporal axis on the basis of visual sensibility and hysteresis. The experimental results illustrate that the proposed model can obtain results in line with subjective scores. It is concluded that the research on HVS characteristics can improve the performance of video QoE prediction method. |
