Optimal End-to-end Distortion Estimation for Drift Management in Scalable Video Coding

This paper is concerned with performance optimization of point-to-point communication of scalable video over lossy networks. The premise of the work is: (i) Decoder drift due to prediction should be controlled but not altogether disallowed; and (ii) Reaping the full benefits of drift management techniques requires accurate estimation of end-to-end distortion. A variant of the recursive optimal per-pixel estimate (ROPE) is developed and embedded within a rate-distortion (RD) optimized mode selection scheme for SNR-scalable video coding. The system allows predicting the current base-layer frame from past enhancement-layer frame data. The ROPE method ensures accurate estimation (at the encoder) of the overall decoder distortion, while explicitly accounting for all factors including quantization, drift, error propagation, packet loss, and concealment. The estimate is used to optimize mode and quantizer step-size selection per macro-block in both the base and enhancement layers. Simulations provide evidence, through substantial performance gains, for the importance of combining drift management with the optimal estimation provided by ROPE.