Research And Implementation Of The FEC Methods For Quality-guaranteed Video Communication

Video stream highly compressed is quite sensitive to channel error because of strong correlation between bits in the stream. Meantime, packet losses can not be avoided in Internet. So it is absolutely necessary to use error control mechanisms in video communication.This thesis is based on network video communication, mainly focusing on Forward Erasure Correction (FEC). Firstly, the background and development of FEC are given, and then RS code, Tornado code LT code and Raptor code are respectively introduced in detail, including the encode and decode method, algorithm complexity and performance. Secondly the Convolutional Tornado code and selective retransmission technology based on FEC are proposed. The bipartite graph of the proposed code is transformed from the group Tornado Code, which can enhance the correlation of packets and avoid the packet losses caused by outburst error. As to selective retransmission based on linear FEC, all of the lost packets can be recovered when received packets can present all source packets linearly. The number of packet retransmission can be calculated by transforming matrix corresponding to bipartite graph when decoder fails to recover all lost packets. Thirdly, packet structure used in retransmission is proposed, in which there are two types of structures involved: information about retransmission packet number and video contents. So that video communication system using proposed packet structure has low complexity and easy implementation.At last, a video communication system based on FEC and retransmission to avoid packet loss is designed. Experiment results show that the Convolutional Tornado code and FEC combined with retransmission can significantly reduce the computational complexity and guarantee the great quality of the reconstruct video, and can be used in the actual video communication system.