Novel error resilient techniques for the robust transport of mpeg-4 video over error-prone networks

K. W. Ng,Bo Yan

2004-01-01

Abstract:In video communications, the transmission of compressed video over error-prone channels requires techniques that are efficient in compressing the video and robust against the channel errors. The application of MPEG-4 can conform to the bandwidth limitation, but the highly compressed bits will be more susceptible to channel errors during transmission. Therefore many researchers have been working on video error resilient and concealment techniques to recover errors caused by channel noises. The MPEG-4 video compression standard has incorporated several error resilience tools including video packet resynchronization, data partitioning, reversible variable length coding and header extension codes. However these tools are not powerful enough for all the error recovery and protection tasks necessary to provide a satisfactory performance for video transmission over noisy networks. Therefore new techniques are required to improve the performance of MPEG-4 video transport over error-prone channels. In this thesis, we present several novel error resilient techniques to strengthen the performance for the robust transport of MPEG-4 video over noisy networks, including error detection and location, spatial and temporal error concealment and channel coding for video data. Firstly, we propose a mode-based error detection (MED) technique, which can detect errors based on the coding mode of the corresponding macroblock (MB). This technique can detect errors, which cannot be detected using traditional methods. Then we present a selective directional spatial interpolation (SDSI) algorithm and a selective motion vector matching (SMVM) algorithm for spatial and temporal error concealments respectively. Next, we propose a joint SDSI and SMVM (JSS) algorithm for error concealment to combine the SDSI and SMVM algorithms together. Finally, for channel coding, we present a mode-based unequal error protection (M-UEP) technique to give more protection to the important bits. Experimental results show that these techniques outperform other existing methods significantly at different channel conditions. In conclusion, these proposed techniques are highly effective and efficient to improve the performance for the robust transport of MPEG-4 video over error-prone channels.