Efficient techniques for two-layer coding of video sequences

Two-layer encoding of packet video may be a useful way to minimize network resource usage while coping with potential packet loss. A two-layer codec architecture modeled after the H.261 RM8 algorithm is proposed, and is shown to be a generalization of the requantization and spectral separation encoding techniques. Experiments were performed using two ten-second video sequences to examine the efficiency of two-layer encoding algorithms. Three techniques: periodic group of blocks (GOB) replenishment, conditional macroblock replenishment, and block error thresholding, are shown to improve the efficiency of two-layer encoding. A hybrid encoding method utilizing both spectral separation and requantization is shown to provide the best results.<
>