| With the rapid advances on computer and network communication technologies, high quality video streaming has become a very popular multimedia application on network nowadays. Video-on-Demand (VOD) can provide subscribers with possibility of watching their desired videos at the time of their choice. Obviously, VOD is widely applied in variety of information service today, and it has a wide application.Constant bit rate (CBR) encoded videos are usually used in most of traditional VOD systems because they are appropriate for transmission. But they suffer from the fluctuant on video quality. Variable bit rate (VBR) coding technique is another encoding strategies. The goal of VBR is to let the quality of encoded video always keep similar. VBR videos have smooth video quality but they suffer from the rate variability. This therefore makes VBR videos inappropriate to be transmitted over the network.In light of the inherent multiple-time-scale burstiness of VBR video, smoothing techniques for reducing the variability are of significant interest. The principle of smoothing is 'work-ahead', which means sending video data ahead of its playback time, and prefetch-ing frames into the clint buffer. By using these techniques in VOD system, The serve can compute a smoothing transmission schedule before transmitting data to client. And a VBR stream can be presented as a series of fixed constant bit rates. In this thesis, we consider the transmission of stored VBR-encoded video from a server to a client across network in VOD system. We proposed a new algorithm, call adaptive-segmented fast smoothing algorithm, which can divide the video stream into segments and determine the transfer rate of segments according the video rate variability. The experiment results showed the proposed algorithm can reduce the peak rate and rate variability significantly. The proposed algorithm are also shown to construct transmission rate-plan with lower computing complexity, as compared to some of existed algorithms .The growth in communication bandwidth may not keep up with the growth in the number of clients if each client is served with a dedicated stream. Periodic Broadcasting schemes are proved to be efficient for transmitting most of the popular videos in VOD system, due to most of subscribers want the few same videos at about a same time. In another word, the program popularity follows K-transformed Zipf-like distribution. Most schemes assume the videos are CBR-encoded, and they can't support VBR video well. A new scheme based Harmonic series scheme, call Equal-length Variable Bandwidth Harmonic Broadcasting(el-VHB), is proposed to support VBR-encoded video in this thesis. We also proposed a general broadcasting model of video segmentation. The workable verification and performance evolution are described in this thesis. The experiment results show that the proposed scheme is outperforms other schemes in maximum bandwidth requirements and maximum waiting time. Furthermore, we enhance the proposed scheme to achieves small bandwidth consumption and buffer demand on the client side.Since the popularity of a video will be changed by many factors, the bandwidth assigned to this video may need to be adjusted to reflect the change, which is called 'channel transition'. It is certainly desirable for a broadcasting scheme to be able to dynamically adjust the bandwidth assigned to each video seamlessly. Only a few of schemes consider this problem, and they can't not support VBR video. In this thesis, we show how to modify the el-VHB scheme to achieve seamless transition. A part of segments will be partitioned or rejoin into new segments, and the relationship between the old and new segments is established. All the segments will be map on the channels according to the broadcasting model. The old and new clients don't suffer from the change of segmentation, and the proposed channel transition scheme is seamless.
|
PDF Full Text Request