Consistent-degradation macroblock grouping for parallel video streams over DiffServ networks

This paper presents a consistent-degradation macroblock grouping scheme for improving loss resilience of parallel video streams over a two-class DiffServ network. By jointly exploiting the H.264 flexible macroblock ordering (FMO) tool, a multi-stream macroblock ordering framework is designed to classify all macroblocks of a super-frame into two categories: important macroblocks as high-reliability traffic class and unimportant macroblocks as best-effort traffic class. To ensure efficiency and fairness of multi-stream macroblock ordering, the compound transmission distortion is introduced to measure both worst-case and average-case loss resilience of parallel video streams. The proposed consistent-degradation macroblock grouping scheme aims to further reduce the compound transmission distortion based on a GOP-level macroblock importance metric, which is adopted to timely estimate the error-propagation effect of macroblock loss on video quality degradation. With finer macroblock classification, the proposed scheme can find out an optimal grouping policy for each super-frame under the given transmission resource constraints. Both the theoretical analysis and simulation results show that the proposed scheme can effectively reduce the compound transmission distortion of parallel video streams over a two-class DiffServ network, and thus provide more consistent loss resilience and better utilization of transmission resources.