An Adaptive Hybrid Mode Decision Scheme for H.264/AVC Video over Unreliable Packet Networks

摘要

In this paper, a hybrid rate-distortion (RD)-based mode switching decision scheme is proposed for robust transmission of H.264/MPEG4-AVC video over unreliable networks. Similar to the recursive optimal pixel estimate (ROPE) algorithm, the proposed encoder recursively estimates the overall distortion of decoder frame reconstruction due to quantization, error propagation, and error concealment. The estimation is then used for switching between intra-coding and inter-coding modes per macro-block. Different from ROPE, our scheme assumes the worst case of the previous frame to make RD optimal mode switching decision for the current frame. To avoid the reduction of coding efficiency, the worst scenario is selectively applied according to the motion activity of the previous frame, which is measured by the potential concealment error. By adopting the notion of concealment candidate image (CCI), we successfully implement the Hybrid-ROPE (HROPE), which finds a good tradeoff between error resilience and coding efficiency. Thus, the HROPE algorithm yields consistent and significant gains over the competing methods. The experimental results under different simulation conditions show that our mode decision scheme has robust error resilient ability and good adaptability. With the advanced feature of H.264, flexible MB ordering (FMO), our scheme can be combined with more existing error concealment techniques. It also attributes to FMO that the ROPE-styled framework in our scheme provides the flexibility of the packet sizes in terms of the maximum transfer unit (MTU) size. Moreover, the hybrid mode decision scheme has low computational complexity and does not require a feedback channel.