A layered multicast packet video system

Abstract

Software based desktop videoconferencing tools are developed to demonstrate techniques necessary for video delivery in heterogeneous packet networks. Using the current network infrastructure and no network resource reservation, a one-to-many implementation is designed around a two-layer pyramidal video coder. During periods of congestion, the network routers give priority to the base layer, which by itself allows reconstruction of reasonable quality video. Receiver feedback is used to lower the output rate of the encoder's low priority pyramidal layer when all receivers are suffering high packet loss. Each of the two layers is transmitted on a separate multicast channel. Under persistent congestion, an individual receiver will discard the low priority pyramidal layer, which allows the network to prune the multicast tree 'd congestion. A new scheme is examined where if the other receivers back and avoi are agreeable, the source will respond to a receiver pruning its pyramidal layer by lowering its rate and allowing the receiver to quickly rejoin the pyramidal layer at a quality level higher than what the high priority base layer can provide by itself. Another new scheme is described where an agent on the receiver's local router provides spare capacity information to assist the receiver in its decision to rejoin the pyramidal layer.