Research On Some Techniques To Improve Streaming Performances Of Progressive Fine Granularity Scalable Video Coding

With Internet and wireless network become much popular and have been applied widely, robust and efficient multimedia services on these platforms show their trends. Among various multimedia services, video transmission involves large amount of data, so efficeint video coding and corresponding transmission techniques are very important. Traditional video coding methods aim to reduce the length of the data. But video transmission over Internet and wireless networks require more features of video coding. Not only the coding performance should be considered, but it is also required that the coded bitstream can provide bandwidth adaptation and error resilience. Fine Granularity Scalable video coding(FGS) is got more concern and has been adopted by MPEG-4 because of its flexible bandwidth adaptation and its compatibility to existing video coding standards. And Progressive Fine Granularity Scalable video coding(PFGS) improves the performance of FGS significantly. However, it is still worse than non-scalable video coding. And wild bandwith adaptation also means that the coded bitstream can not always work optimally at all bit-rates. To improve the performance of PFGS, we present some schemes. The contributions of this paper are as following:1. A high efficient PFGS bitstream swiching scheme is proposed. The basic idea is to switch several PFGS enhancement layers' bitstreams dynamically according to the bandwith change to improve the streaming perforamce, while the base layer remains the same. To generate these enhancement layers' bitstreams efficiently, we design a PFGS encoder that can generate several PFGS enhancement layer's bitstreams and a base layer's bitstream simultaneously. We also provide a rate-distortion based method to evaluate different bitstreams' performances at different bit-rates. Experimental results show the efficiency of the proposed switching scheme.2. A high efficient PFGS transcoding scheme is presented. The basic idea is to change the enhancement bitstream transmitted to improve the performance while the base layer remains the same. The difference between the PFGS transcoding scheme and the PFGS switching scheme is that the former does not need multiple bitstreams, thus it means that much encoding time and storage space can be saved. But the PFGS transcoding needs more computation during transmission. We design simple PFGS transcoder and experimental results show its efficiency.3. We propose rate allocation schemes for Progressive Fine Granularity scalable video coding with temporal scalabilities(PFGST). We analyze the different properties of different frames, then a reference-first rate allocation scheme is presented. And further we propose a rate-distortion based rate allocation scheme, which can give optimal results. To satisfy the buffer constrain in practice, a sliding window based rate allocation scheme is also proposed. Experimental results show that our rate allocation schemes can significantly improve the streaming performance of PFGST. The gain can be up to 7dB.