Two-scale transmission of DCT-coded video over lossy packet networks

This paper discusses the development and results of a simulation package for two-scale transmission of discrete cosine transform (DCT)-coded video over lossy packet networks. The considered scalability of coded video sequences is based on variations of data partitioning and requantization techniques. These variations contemplate situations in which the reference signal used in predicted (P-type) or interpolated (B-type) frames is based on the full coded signal, or alternatively, based only on the high-priority signal. Also considered are variations of data partitioning with a fixed number of spectral coefficients. The signals relating to the two scales are to be transmitted over a channel formed by a high-priority link, with no loss of cells, and a low-priority, lossy link. The results of transmission over packet networks with various levels of cell loss rate are presented and a few conclusions are drawn.