The Video Quality Assessment Model Based On Streaming Media

With the development and popularization of streaming media technology，network video has become a mainstream multimedia carrier. In the process ofstreaming transmission, we always handle the video by compression, transmission,decompression and other operations. Those operations would decrease the quality ofthe videos, which, however, determines the service quality of the streaming mediasystem. As a consequence, a growing number of the streaming media operators aredemanding intuitive and accurate video quality evaluation model based on streamingmedia, to get the quality value of network video. It would help the operators to graspthe current performance of streaming media, select the appropriate streaming mediaserver, and promote the streaming media technology. Therefore, the study of evalua-tion quality of streaming media video is very significance.This thesis focuses on the non-reference video quality evaluation, because it canwork well without the reference of the original video. It is suitable for batch pro-cessing. We can get the quality result in time by this method and use our assessmentmethod in the streaming media system.Firstly, we put forward the blurring effect assessment based on the linear predic-tion error. We use the least square method to obtain the linear combination of the im-age blocks which helps us compute the difference of the predicted values and theoriginal values of pixels. We use this difference as the characteristic value of blurringassessment.Secondly, we improve the method of the blocking effect assessment based on the periodic block detection, as well as the noise effect assessment based on the wavelettransform. For the blocking effect assessment, we increase three constraints to betterdistinguish the difference between image texture and blocking texture, to improve thereliability of block effect assessment. For the noise effect assessment, we use threewavelets with floating threshold processing the image, to improve the effectiveness ofnoise effect assessment.Thirdly, we use the machine learning method integrating these three kinds ofdistortion effect assessment, and give the video quality comprehensive assessment.We train the database with the subjective quality value by machine learning method toget some features. Then we use these features to integrate these three kinds of distor-tion effect measurement methods. Finally, we would give a method of video qualitycomprehensive assessment.Fourthly, we implement the video quality assessment model based on thestreaming media and video assessment plugin on the Google Chrome Web Apps. Weuse relevant technologies, such as the streaming media technology, the network videocrawler technology, the key frame extraction for video summarization, and the videoquality comprehensive assessment. Then we give the video quality assessment modelbased on streaming media technology. According to the result of video search, we findthat the video quality assessment plugin on the Google Chrome Web Apps, read thevideo quality value in real time from the database of the assessment model. Also itprovides the video quality to user through the web page.