Jointly optimal reference frame & quality of service selection for H.261 video coding over lossy networks

In new video coding standards such as H.26L, predicted frame has the flexibility to select its reference frame from a number of previous frames for motion prediction. In this paper, we propose an optimization algorithm using dynamic programming that jointly exploits this flexibility with available network QoS for optimal streaming performance. A rounding technique is employed to scale the complexity of the algorithm down at the expense of gracefully degrading solution quality. Results show significant streaming quality improvement over an ad-hoc scheme.