RD-Based Adaptive UEP for H.264 Video Transmission in Wireless Networks

Video streaming is challenging in wireless networks due to the high loss rate and limited bandwidth. In video sequences, different part has different importance for reconstruction. Based on H.264 error resilient tool of datapartitioning, we apply UEP (Unequal Error Protection) to different data partitions. RS (Reed Solomon) coding is used for data partition A, whereas simple repeat-packet coding for data partition B and C. An RD (Rate-Distortion) model is then established to optimize the bit rates distributed among different data partitions and parity codes, in order to minimize the end-to-end distortion. The model accuracy is verified. Based on the RD model, an adaptive UEP framework is proposed, under which the strengthes of RS codes and repeat-packet codes, and the quantization parameters of the source codes are determined dynamically according to the network conditions in terms of available bandwidth and packet loss rate. The extensive experiments show that the proposed scheme can adapt to the network conditions very well, graceful reconstruction quality degradation can be achieved against a large range of packet loss rate and available bandwidth.