Performance and statistical analysis of two-layer MPEG coding algorithms for ATM networks

Two-layer video encoding algorithms for asynchronous transfer mode (ATM) networks are examined. Three encoding schemes are considered to prioritise the data using either requantisation or spectral separation or a proposed hybrid combination of both methods, The hybrid coder assigns components of the video data stream to one of two priority layers based on the macroblock type and the algorithm structure, Statistics of cells-per-frame processes generated by the proposed hybrid encoder such as the cell rate distribution and autocorrelation function are evaluated, It is observed that the processes are nonstationary and a gamma density function is a suitable model for the cell rate distribution in both layers. Two multiplicative ARIMA models are derived to represent the nonstationary processes mathematically. The statistics measured from a sample video sequence compare favourably with those obtained by this model, Statistics gathered from these prioritisation schemes can be useful for network design and resource management. Simulations are performed using a 5-second video sequence.