An adaptive key-frame reference picture selection algorithm for video transmission via error prone networks

When compressed video is transmitted via error prone network, it can suffer from severe degradation. To combat the effect of transmission errors, a novel adaptive key-frame reference picture selection (AKRPS) algorithm is proposed. In this algorithm, key-frames are predefined according to the round-trip delay, and the latest key-frame received ACK signal is used as the reference picture of key-frame while the other frames is encoded normally, which can prevent the temporal error propagation efficiently with keeping high coding efficiency. The required reference picture memory for AKRPS is less than 3 frames independent of the round-trip delay. A frame-layer rate control scheme is also proposed for smoothing the video quality. Simulation results, with different video sequence and H.263+ codec, show that the KRPS can achieve high error resilience at packet error rates (PER) ranging from 1% to 10% in both single-user and multi-user applications.