VIDEO OVER ETHERNET

Abstract

Vast improvements in network capacity and data compression/decompression techniques have now made it possible to provide multimedia on demand across various types of computer networks. Many applications providing video and audio data over computer networks have been developed. Examples include the network video (NV) tool [ 1 ] and the INRIA Videoconferencing System (IVS) [ 2 ] currently available from the Internet. The number of such application is expected to increase significantly over the next few years and indeed, the whole computer industry is currently betting heavily on the future success of networked multimedia. For the video application, the ISO 11172 or MPEG1 video coding standard has been developed by the Motion Picture Expert Group (MPEG). The developed standard allows the transmission of broadcast quality images at a data rate of approximately l to 1.5 megabits per second and support features such as forward play, fast-forward play, etc. In this project, an experimental networked video retrieval and playback application has been developed over an Ethernet network to allow multiple clients to access common pool of MPEG video files stored at a remote server. The application allows a user to control the playback of the video similar to that allowed by a video cassette player (VCR). The MPEG video files are transferred on a picture-by-picture basis and decoded at the client before playback using software decoding. Three different algorithms for controlling the flow of the video pictures from the server to the clients have been developed. The first algorithm utilizes a single picture buffer at the client to store the picture before playback. The second algorithm utilizes a small buffer that is sufficiently big to pre-store the pictures before playback. Both the first and second methods perform feedback to the server to control the flow of video pictures. The third algorithm is similar to the second except that no feedback is provided by the client for video picture flow control. Tests conducted c11 the three different algorithms show that the second algorithm produces the best result in terms of the playback rate against requested rate, the number of buffer overflows and pictures dropped, and the subjective test on the quality of video.