Research On The Key Error Control Techniques Of Video Transmission Over The Internet

In recent years, the Internet based multimedia applications, especially the video applications, are getting wider and wider. Video compression and transmission have gained broad attention and become one of the most challenging problems. Traditional video schemes are based on the ideal environment without error, and are designed to improve the compression ratio. They are not suitable for the packet-switch based, best effort, and unreliable Internet. This contradiction makes high-quality video transmission a challenge to be resolved.Currently, all the video coding standards, such as MPEG-4, adopt the variable length coding and motion prediction techniques to improve the compression ratio. However, in error prone environment, the network transmission error, such as packet loss, may lead to the loss of video data at the receiver, which means the loss of many video blocks, and possibly video frames. This kind of error will lead to the error propagation in the temporal and spatial domain. Therefore, it is necessary to introduce error control techniques during the process of video transmission.For Internet, the phenomena such as bandwidth variation, packet loss and bit transmission error are inevitable. All kinds of error control techniques are the indispensable components of video transmission. Error control techniques aim to effectively improve the quality of output video at the receiver. They can function at any step of the encoding and decoding process of source and channel. Therefore, the range of error control techniques is quite broad, and can be roughly divided into three categories: forward error-resilient encoding, post-processing error concealment, and interactive error control in codec. This thesis researches on the key techniques of the first two categories, analyzes the limitations of existed algorithms, and then proposes corresponding solutions and improved algorithms. The main work of this thesis is summarized as follows.Firstly, the strategy of multiple descriptions coding (MDC) for DCT at the encoder is discussed. In order to improve the robustness of video transmission, the original video is interleaved into multiple descriptions before encoding. Optimized reconstruction-based DCT (ORB-DCT) scheme aims to minimize the errors between the reconstruction video and the original video when one description is received. An improved ORB-DCT (IORB-DCT) scheme is proposed based on the correlation between neighboring pixels. The improved algorithm can enhance the pixel correlation in description, and thus provide more information during the reconstruction at the decoder.Secondly, the temporal domain error controlling techniques for P frame is investigated. Under certain condition of packet loss in Internet, a packet loss usually leads to the loss of P frame. Therefore, to recovery the loss of P frame, the motion information of multiple previous motion compensation frames are utilized. After discussing the error concealment techniques using dual forward predictive motion compensation, multiple motion vectors are utilized to conceal the loss by considering the buffer capabilities at the decoder and the relationship among motion vectors.Thirdly, a novel scattered packet algorithm based on multiple jumpy scans is proposed by fusing the advantages of traditional raster-order scan and recent scattered slices. Packet algorithm is one of the most important and effective techniques to resist errors. The proposed approach is to overcome the limitation of scattered packet. Because it scans the scattered macroblocks several times in certain order, it can overcome the limitation of macroblock's fixed numbers in scattered slices, and optimized, bit based scattered packet is realized.Finally, since packet loss and bandwidth variation are two major factors, which affect the quality of video, Unequal Error Protection (UEP) and MDC based fine granularity scalability (FGS) encoding scheme is proposed. The bitstream produced by FGS can well adapt the different bandwidth of Internet, and MDC can deal with the packet loss. Because the bitstream of base layer and enhanced layer are of different importance, and different bit planes of enhanced layer have different importance, UEP protection is adopted for every layer. After analyzing the characteristic of FGS bitstream, especially the characteristics of the bit planes of enhanced layer, MDC is performed on base layer and partial enhanced layer respectively to ensure the basic video information and the correct transmission the enhanced information, which can enhance the video quality greatly.