文档下载 文档下载导航
设为首页 | 加入收藏
搜索 请输入内容:  
 导航当前位置: 文档下载 > 所有分类 > Prediction of Cell Losses for VBR MPEG Encoders
侵权投诉

Prediction of Cell Losses for VBR MPEG Encoders

Abstract: In this paper modelling and queuing analysis of variable-bit-rate (VBR) MPEG coded video sources in B-ISDN networks based on ATM are studied. Modelling is performed by examining a low pass version of the aggregate MPEG sequence, capable of captur

Prediction of Cell Losses for VBR MPEG

Encoders

Anastasios Doulamis, Nikolaos Doulamis and Stefanos Kollias

National Technical University of Athens

Dept. of Electrical and Computer Engineering

email: adoulam@image.ntua.gr

Abstract: In this paper modelling and queuing analysis of variable-bit-rate (VBR) MPEG coded videosources in B-ISDN networks based on ATM are studied. Modelling is performed by examining a low passversion of the aggregate MPEG sequence, capable of capturing the video activity. A discrete state continuoustime Markov chain is then proposed to approximate the aggregate MPEG traffic. The previous model is used,combined with flow equivalent queuing analysis, to obtain analytical solution of cell losses. Experimentalresults are presented indicating that the proposed model is a good predictor of network resources.1. Introduction

No doubt, digital video is expected to be a major traffic component over B-ISDN networks.However, uncompressed video information is hardly transmitted, even over high-speed cables, dueto large bandwidth requirements and the increasing demands for video transmission. Several codingalgorithms have been proposed in the literature for achieving efficient video compression whilesimultaneously maintaining an acceptable picture quality. Recent efforts towards this goal havestimulated the MPEG-1/2 coding standards, which have been developed for storage of compressedvideo and TV transmission respectively and now appear in commercial exploitation.

Technological progress both in field of electronics and optics, as well as the requirement for aflexible telecommunication network design, able to integrate various services of different kind, ledthe CCITT to adopt the Asynchronous Transfer Mode (ATM) principle. However, this moderequires the development of effective congestion control schemes which are responsible formaintaining an acceptable quality of service (QoS) level that is derivable by the network. Accordingto the available network resources, these schemes ensure that the negotiated traffic characteristicparameters are not exceeded. To determine which congestion scheme should be used, modelling oftransmitted information is required.

In real time video or multimedia applications, the network should guarantee a minimum level ofpicture distortion. However, if a tolerated picture degradation level is given, encoder rate will betime varying, according to the scene complexity (variable bit rate-VBR-coding mode). The maindifficulty of transmitting VBR coded video streams through communication lines, which are ofconstant rate, is that they also require time varying bandwidth. To “tailor” VBR encoders into achannel of speed lower than the maximum coding rate, buffers are used which absorbs statisticalpeaks of the traffic by temporarily storing video information. In this case if the amount of dataexceed the buffer size, then losses occur.

Several video models for VBR coded sources have been proposed in the literature. The firstpioneer attempts were by Haskell and Limp who proposed and simulated statistical multiplexingvideo for picturephone encoders [6,9]. In [10] a single video stream of a telephone scene is analysedand two models are shown to fit well the experimental data. A more accurate model has beenproposed in [7] for VBR teleconference sequences. Scene modelling is performed by Frater et al.

[5] for video films whose the autocorrelation functions indicate a long-range dependence.Since traffic models depend on video statistical properties and characteristics, theaforementioned approaches cannot directly be applied to MPEG coded video sources. Recently,some works, mainly dealing with MPEG statistical properties, have been reported. An approach toefficient modelling of MPEG coded video sources has been proposed in one of our earlier work [4]by examining MPEG streams at three different layers. However, in this work computer simulationis required for estimating the network resources since the involved models are not susceptible ofanalytical solution. In this paper, we perform queuing analysis, based on Markov models, and ananalytical solution of estimating the appropriate buffer size and cell losses is derived. The proposedmethodology provides accurate prediction of the actual traffic for point to point connections.

第1页

你可能喜欢

热门文档

相关文档

站点地图 | 文档上传 | 侵权投诉 | 手机版
本站所有资源均来自互联网,本站只负责收集和整理,均不承担任何法律责任,如有侵权等其它行为请联系我们.
文档下载 Copyright 2013 doc.guandang.net All Rights Reserved.  -  email