Low-complexity rate-distortion optimized video streaming

This paper proposes two techniques for low-complexity rate-distortion (R-D) optimized streaming of packetized video. These techniques enable computing packet transmission schedules which satisfy a constraint on the average transmission rate while at the same time minimizing the average end-to-end distortion. Optimized packet schedules are computed with considerably lower complexity as compared to conventional algorithms for R-D optimized streaming, which makes these techniques suitable for on-line optimized streaming. Simulation experiments examine the performance of the proposed techniques using JVT/H.264 encoded video sequences and previous frame error concealment. The two techniques demonstrate substantial performance gains of 2-8 dB over a conventional streaming system that is not R-D optimized, which corresponds to a significant fraction of the gain achieved by current (high-complexity) R-D optimized schemes. Furthermore, this performance improvement is achieved with a complexity comparable to that of the conventional, non-R-D optimized, system.