Research On Technologies Of Adaptive Video Streaming Based On Rate-distortion

With the continuous development of network technology, the business of video streaming is growing at a remarkable rate. Multimedia data, especially video data, is different from the other data, which has the characteristics of large data, long duration and responsive to propagation delay. These characteristics bring enormous challenge to multimedia techniques and network technology. As the speedy development of coding techniques, video data has high compression ratio. However, in order to guarantee the quality of video playback, a large amount of network bandwidth is still required when transmitting video streaming. Moreover, high compression ratio can give rise to sensitization for the lossy packets over the network. When current bandwidth is too small to transmit video streaming, the network node will drop packets randomly. The result of dropping packets randomly is decreasing video quality.In view of the above questions, this paper research into the algorithm of scheduling video streaming and improve the current packet dropping algorithm based on rate-distortion optimized system framework. The contributions of this thesis include:(1) We introduce the characteristic and requirements of video and the challenge for transmitting video streaming. On the one hand, video streaming has the characteristics of large data and longest-running. As the speedy development of coding techniques, video data has high compression ratio, which leads to the decrease of video quality in the lossy network. On the other hand, current network technology still has the defect in bandwidth, error code, network heterogeneity and so on. Above questions are important for researching into transmitting video streaming. We introduce and analyze the limitations of conventional self-adapting technology for transmitting video streaming. Specially, it emphasizes the self-adapting technology for transmitting video streaming based on rate-distortion optimized.(2) It elaborates on streaming media system framework and rate-distortion side information about video. The characteristics of video streaming include video motion intensity and video frame side information using a dynamic frame concealment pattern. Firstly, using a dynamic frame concealment pattern helps the accurate calculations of the importance of frames and the reconstructed video quality is improved. Secondly, we analyze the calculation method of scene motion intensity and illustrate the relationship between video quality and scene motion intensity.(3) A content-aware distortion-fair algorithm for H.264 data partitioned video streaming is proposed. When video clips are competing for the limited network bandwidth resources, by the bandwidth fair sharing, they may not have equal video quality and happiness. Therefore, the algorithm supports content-aware for videos and according to the scene motion intensity drop the packets with limited bandwidth. It will be formalized as an unconstrained nonlinear optimization problem. We use a dynamic frame concealment pattern to improve the reconstructed video quality. To further cope with frame dropping scheme's shortcoming that only temporal scalability is provided, the algorithm introduce H.264/AVC data partitioned technology. By this technology, every frame is divided into three parts of different priority. Simulation results show that, compared with other dropping packets algorithm, the proposed algorithm can improve the overall reconstructed video quality.(4)Adopting the JMF&RTP/RTCP protocol in prototype system about video frame dropping, the video is transmitted and replayed in real time. Under the same conditions, we test the content-aware dropping frame model and random dropping frame model. The test result shows the content-aware distortion-fair algorithm is feasible.