Analysis, modelling and performance prediction of digital video statistical multiplexing

Statistical multiplexing is rapidly becoming an integral part of many digital video broadcasting systems. These systems are designed to ensure optimum usage of channel capacity while providing consistent picture quality through the use of automatic bit-rate allocation. The problem of determining the optimal bit-rate to give each channel is no longer an issue for the broadcaster, except to determine the total bandwidth to be allocated to the statistical multiplexer. Broadcasters familiar with their programme mix (e.g. sports, news, etc.), wish to know the implications on picture quality, given parameters such as the number of channels of each type and the total bandwidth available to the system. Furthermore, large broadcasting systems present the opportunity to `mix and match´ the source material, to optimise performance and ensure sustainable picture quality. This paper investigates the statistics of existing satellite broadcasting channels and demonstrates that the various types of programmes exhibit measurable statistical differences in a consistent manner. By taking important coding parameters such as coded picture resolution into account, the paper derives accurate models to predict the performance of such systems through characterisation of the video sources to be multiplexed