A novel framework for robust video streaming based on H.264/AVC MGS coding and unequal error protection

We present a novel framework to provide robust video streaming service over time-varying error-prone network. The scheme is based on the medium granularity scalability (MGS) video coding of the H.264/AVC standard, which adopts a hierarchical prediction structure for the group-of-pictures (GOP). We determine the optimal allocation of protection strength for different network abstraction layer (NAL) units according to their individual importance to the end-to-end video quality. To analyse the importance of the NAL units, we emulate the error concealment if one frame is considered as lost and take into account the propagation distortion within the GOP. An efficient algorithm is proposed to account for the non-convex rate-distortion characteristics associated with the NAL units in the hierarchical GOP. With this framework, we can provide robust video streaming for the range of packet loss rates from 0% to 40% with about 30% additional channel bit-rate for the channel coding. The simulation results demonstrate high flexibility and efficiency of the proposed framework, which can effectively prevent frequent loss of frames.