Two-layer video coding and priority statistical multiplexing over ATM networks

Abstract

A two-layer video coding technique is developed. The video signals, after discrete cosine transformation (DCT), are coded via a hybrid compression scheme of a vector quantization technique (VQ) and an entropy coder at the primary layer, while the distortion between the original video signals and the decoded output of the VQ is coded by the means of an entropy coder which is compatible with the MPEG standard at the secondary layer. The peak traffic generated by an intra-frame at the primary layer is quasi-constant. Therefore, bandwidth can be easily allocated for the traffic at this layer. Since the traffic at the primary layer is a small fraction of the overall traffic, there is no cell loss due to network congestion at this layer. The traffic at the secondary layer is adapted to the congestion in the network via a novel adaptive bit rate control scheme based on the occupancy of the multiplexing buffer at end of each frame interval. With this scheme, the cell-loss-rate (CLR) at the secondary layer is, and the overall quality of service (QoS) is improved. Actual video sequences are used to validate the two-layer video coder and the adaptive bit rate control scheme. The CLR reduces in magnitude under high network utilization when our bit rate control scheme is applied. The experimental results are compared with the analytical calculation with the assumption of an M/M/1/N queuing model. We find that the CLR from the numerical simulation is close to that of the analytical calculation especially when the network is highly utilized.