Wireless Environment Error Resilient H.264-based Technology Research

With the development of mobile communications network, the bandwidth and data transfer rate greatly increase, which make the video transmission over wireless channel possible. However, video data is very large. Meanwhile the payment for the wireless bearer services is proportional to the transmitted data volume on it. Then a codec with high compression efficiency is desired and H.264 came into being. But just because of the high compression ratio which removes as much redundancy as possible, coupled with the error-prone wireless transmission channel, which make the encoded bit stream extremely fragile to various errors. Therefore, research on the error control and concealment of H.264-based wireless applications has a realistic significance and urgency.The dissertation first presents some key technologies during the encoding process. Then various techniques on error-control, error-detection and error-concealment are reviewed and an in-depth analysis of their applications is also explored. Moreover, the dissertation expatiates the error control methodologies, such as Data Partition,Flexible Macroblock Ordering(FMO), which are newly introduced in H.264 standards. Besides, some relative information on content-based concealment is also introduced.On the basis of the work above, this dissertation makes an improvement to the error control program based on picture-partition and FMO to solve the defect of high bit-rate and the improper foreground segmentation. Afterwards, MobileIP, the common test condition provided by JVT, is used to emulate the RTP/IP-based transmission on 3GPP/3GPP2 network. And experiments are made out using various video sequences complying with the proposal of the tool. Then a comparison is performed between the original scheme and the improved one with the aspects of compression ratio, decoding time, PSNR and subjective quality assertion. Results show that the proposed improvement can effectively reduce the bit-rates and decoding time. Also some rising of PSNR and better visual perception are achieved to most of the video sequences.As for the variety of different error resilience tools, a developing trend is to self-adaptively select the appropriate one according to the specific application context, or combine some of them. The content-based error concealment, which outperforms the traditional ones with the ability to obtain closer subject perception, is gaining more and more attention.