Research On No-Reference Video Quality Assessment

In recent years, with the rapid development of wireless technology and broadband transmission technology, many multimedia applications, such as video conferencing, video-on-demand and video chatting, are becoming more and more common. In the research report about Visual Networking Index, Cisco forecasts that the global video-based mobile data traffic will account for66%of the total flow by2014. In order to improve the quality of video and user experience, telecom operators and service providers need to predict the video quality on user terminal, monitor performance of the transmission system, find out cause of video distortion and improve the transmission condition timely. This means that the research on Video Quality Assessment which related to user experience shows an unprecedented importance.Since video encoding and video transmission environment are non-ideal, video will be distorted in the process of capture, processing, transmission and reconstruction, and user received video is not clear. No-reference video quality assessment based on a variety of distortion type presents a better evaluation result for different video content and different distortion type and becomes one of the most practical quality assessment methods. This article studies the No-Reference Video Quality Assessment methods and puts forward the following research results:(1) Accurately detect the block effect, blurry and jitter distortion of image. The frequency-domain method is used to detect block effect. The comprehensive method of frequency-domain and time-domain algorithm can effectively detect the blurry distortion. In this paper, the distortion introduced by network packet loss is called jitter distortion. The proposed algorithm can better detect jitter and forecasts the network packet loss in turn.(2) Propose two new combined algorithms. The SSIM algorithm is used to calculate image quality score. We analysis the advantages and disadvantages of existing video combined algorithms, then put forward two new combined methods. The proposed time-domain combined methods can provide a better result.(3) Analyses how mixed distortions impact the video quality. Analysis interaction relations among block effect, blur distortion and jitter distortion, and conclude a relational model which will be used among the single frame image quality score. Multi-parameter detection methods make up the deficiency of single distortion characteristic detection methods and improve the evaluation quality and accuracy.