Study On Error Resilient Video Coding Method Based On H.264/AVC

With the rapid development of Internet, wireless communication, and interactive multimedia, low bit-rate video reliable communication is becoming an important issue in the research area of video communication today.Today's network technologies, including the Internet and wireless communication, can not provide a guaranteed quality of service (QoS), so the video data packet will be lost or corrupted in the transmission process. In such situation, the error-free delivery of data packets can only be achieved by allowing retransmission of lost or damaged packets, through the mechanisms such as automatic repeat request (ARQ). However, such retransmission mechanism will incur delays which are unacceptable for real time applications. Therefore, it is important to design a video encoding/decoding which can make the compressed video stream resilient to transmission errors.H.264/AVC is the new generation of video coding standard out of the joint effot of ISO MPEG and ITU-T VCEG. By using multiple reference picture motion compensation, variable block-size motion compensation with small size, 4x4 block size transform and directional spatial prediction for intra coding etc., H.264/AVC improved the coding efficiency significantlly. Howerver, due to the hybrid coding which H.264/AVC adopted, the compressed video stream is more sensitive to transmission errors. Moreover H.264/AVC used several new technologies, so it became an important research topic to design error resilient video coding technologies for H.264/AVC video coding standard.This dissertation conducted the research focusing on the error resilient video coding for H.264/AVC video coding standard. Our reserch includes not only implementing and analizing interactive error resilient video coding, but also proposing and implementing, by ourselves, the mode adaptive temporal error concealment algorithm based on backward motion tracking. The encoder and decoder should be designed for implementing the interactive video coding, which need the cooperation between the encoder and the decoder. On the other hand, the decoder shoud be designed to implement the error concealment. In summary, the content of our work is organized as follows:Firstly, we conducted research focusing on analyzing the effectiveness of an interactive error resilient video coding technology, feedback based reference picture selection (RPS). We implemented the error resilient video coding based on feedback RPS, making use of the multiple reference picture motion compensation technique. Then we proved that the error resilient video coding and the source coding can be achieved simultaneously by combining RPS and the multiple reference picture motion compensation in H.264/AVC, to take both error resilience and high coding efficiency into account. When presenting the process of implementing the RPS algorithm, the syntax of H.264/AVC is analised, firstlly. Secondlly, implementation of the RPS, which is compatible with H.264/AVC video coding standard, is presented in detail. Finally, we conducted the simulations according to common conditions for error resilient testing JVT-P206, which established by JVT and compared the simulation results with another error resilient video coding technology, random intra refresh (RIR). The simulation results show that the feedback based error resilient video coding technology has better performance than the RIR in H.264/AVC. With 3% packet loss rate and the 6 frames feedback delay, the average Y-PSNR of different reconstructed sequences has 2-7 dB gains.Secondly, we proposed mode adaptive temporal error concealment algorithm based on backward motion tracking, effective using of the variable block-size motion compensation with small block sizes. The error concealment technology proposed by us prevents multiusing motion vector estimation to solve the problem of the error propagation. Moreover in order to adapt to the variable block-size motion compensation technology used by H.264/AVC, the mode adaptive error concealment algorithm estimates the corrupted macroblock mode based on the backward motion tracking, and then conceals the error mode adaptivelly. Finally, we proved the effectiveness of proposed algorithm by presenting experimental results. With packet loss rate 3% and the slice size 100 bytes, the average Y-PSNR of reconstructed video sequences coastgard and foreman have 0.6dB and 0.2dB gains respectively at NACK mode.