Research On Key Techniques Of Internet-oriented Video Transmission

With the development of Internet and video codec technology, the video transmission plays an increasingly important role in the data transmission service of Internet. Since the characteristics of video, including huge amount of data, high encoding and decoding complexity, and strong dependencies among each data unit, video is extremely sensitive to QoS parameters such as packet loss and delay. While due to the rapid expansion of Internet and growing mobile devices accessing it,especially under the tendency of integration of mobile network and Internet, the quality of service is further reducing,which lead video transmission over Internet to face problems such as data error,distortion,and interrupt. Therefore, centering on the study on the real-time capability and stability of video transmission, and the quality of video on receiver, has become an important topic in the data communication and video codec research. As the streaming business represented by video become increasingly prevalent on the Internet, the research of related techniques not only possess prominents academic meanings,but also high value in applications.Based on the existing research achievements at home and abroad, this dissertation makes the research of key techniques of Internet-oriented video transmission. The main works of this dissertation are as follows:(1) Error correction technique for channel codingDue to the video transmission over Internet is prone to packet loss, a predicting packet loss rate based error correction method in video transmission was proposed. First, the relationship between packet loss rate and video distortion is analyzed. Then the packet loss rate is quantified by FCM, and the Baum-Welch algorithm is used to train hidden Markov model (HMM). The solution for HMM is calculated by employ Viterbi algorithm.Therefore, the packet loss rate in next peorid is obtained. According to the statistics, the forward error correction (FEC) or automatic repeat request (ARQ) was adaptively selected to perform error correction on the video data. The Simulation experiment reveals that the proposed method could achieves better subjective and objective video quality and effectively reduce the average delay of video transmission, when compare with the methods HARQ and LIVA.(2) Error correction techniques for scalable video coding In order to improve the quality of video at receiver, an efficient adaptive genetic algorithm based error correction method for scalable video coding(SVC) over Internet was proposed. First, a scheme to encapsulate network abstract layer units according to the feature of the head information of packet was introduced. Secondly, according to the different functions of layers in SVC, the problem of pair codes assignment were transformed into problem of multi-constraint optimization on the basis of distortion modeling, which could be transformed into unconstrained objective by exploiting penalty function. Finally, the adaptive genetic algorithm is employed to obtain the optimized FEC data allocation scheme,and the minimum average distortion is obtained. Simulation results reveal that compared with the typical methods LW-EZEP and GOP-UEP, the proposed method achieves better objective video quality and gives substantial improvement for the decoding speed.(3) Unicast Congestion Control based Video Transmission TechniqueDue to the existing unicast congestion control methods have some problems in the reliability and distortion, a Semi-Markov Decision Process based unicast congestion control method was proposed. First, in order to achieve the improvement of real time, an online operation no-reference video quality assessment scheme was presented. Secondly, in order to minimize the distortion at receiver, the video transmission distortion is analyzed and the unicast congestion control problem was formulated as a Semi-Markov Decision Process according to the feedback of video quality assessment. Therefore, a calculation method with high real time is presented to solve the model, and the optimal strategy for parameters of congestion control is obtained. Simulation results reveal that the proposed method obtains well TCP-Friendliness ratio, and achieves better subjective and objective video quality, when compare with the typical unicast congestion control methods MTCC and TCP reno.(4) Multicast Congestion Control based Video Transmission TechniqueIn order to overcome the limitation such as lower bandwidth utilization and high average packet delay, a multi-rate multicast congestion control method based on simulated annealing was proposed. Fist, under the premise of maximizing the require rate of receivers, the multi-rate multicast congestion control problem for video transmission is formulated as a nonlinear optimization problem, then the simulated annealing algorithm was employed to obtain globally optimal solution and improve bandwidth utilization effectively. Secondly, due to loss data recovery was performed by proxy node, the frame availability and reliability of transmission is improved. And the scheme of layer adjustment is presented to reduce average delay. The results of simulation demonstrates that the proposed method not only achieves better bandwidth utilization and throughput, but also the scalability and friendliness, obtained improved quality of video, when compare with the conventional multi-rate multicast congestion control method TFMCC.(5) Rate Control based Video Transmission TechniqueDue to the conventional rate control algorithm for the latest video coding standard H.264/AVC without take into account the human visual perception and result in low accuracy and video quality fluctuation. A visual perception based rate control method for H.264/AVC is proposed. First, the method to detect the scene change and characteristics of local motion is presented, then the activity of frame is defined. Secondly, in order to conform to characters of visual perception, the pixel level based Just Noticeable Distortion(JND) model is obtained by using visual perception. Therefore, the bit allocation to frame is performed according to the JND of frame. Finally, in order to effectively eliminate the time and space redundancy,the structural similarity index as the quality metric for rate-distortion modeling is proposed and the optimum bit allocation is performed in basic unit layer. Experimental results demonstrates that the proposed rate control method not only improved the video quality, but also significantly increased the control precision and raised stability of video,when compare with the typical rate control methods VPBRC and JVT-W042.At the end of this dissertation, the main research is summarized. It makes out the main research achievements and innovations, and also points out the problems and issues which need to further research.