End-to-end distortion estimation for RD-based robust delivery of pre-compressed video

Applications where packetized video is streamed over the Internet, must be designed to achieve robustness to packet loss as well as compression efficiency. Whenever possible, the ideal solution to this problem is to jointly optimize the adaptation of the compression and error protection strategies to the network status, so as to minimize the expected end-to-end distortion of reconstructed video at the receiver. However, in the case of pre-compressed video streaming, compression is performed without knowledge of the network condition, and conversely the delivery is performed without access to the original signal. It is hence difficult for the transmitter to estimate and minimize the end-to-end distortion during delivery. This paper addresses this problem by deriving an algorithm which enables the transmitter, or other intermediate nodes, to estimate the overall end-to-end distortion while delivering pre-compressed video. This estimate fully accounts for the effects of (prior) quantization, packet loss and error propagation, as well as error concealment. The accuracy of the estimate is demonstrated by simulation results. The algorithm requires storage of minimal side-information that is computed during compression. The algorithm is of low complexity, and is applicable to virtually all coding techniques, including the standard (predictive) video coders. The paper also discusses the use of this estimate to adapt a variety of packet-loss resilience techniques for pre-compressed video streaming. The considerable potential gains of this approach are illustrated via the example of an FEC-based streaming video system.