Research On Video Error-Resilient Techniques For MPEG

With the fast development of telecommunication technologies, video communication has been widely applied and the related techniques have been developing quickly. Video carries a huge amount of information and is therefore often compressed before transmission. However, as real-life communication channels are hardly error-free, the data transmitted through such channels is inevitably corrupted, especially when wireless communication is involved. Block-based video coding systems are vulnerable to transmission impairment. One single bit error often results in the loss of a whole block and its consecutive blocks. What's worse, the error may propagate to the several consecutive frames through temporal prediction. All these vulnerability to error is caused by removal of natural redundancy during compression. Therefore, proper techniques for error resilience must be applied in order to achieve acceptable quality of restored video.The research in this dissertation mainly focuses on the techniques introduced in video encoders and those at decoder endAs the prerequisite to error processing, video error detection plays an important role in real-life applications. We propose a two-stage error detection method for MPEG-2 which detects errors both in terms of bitstream syntax and spatial and temporal continuity. Apart from detecting erroneous slices via syntax-based errors, this algorithm recognizes special error patterns caused by erroneous DCT coefficients, which is a clue to erroneous slices ignored by syntax-based method. When searching for the first erroneous macroblock in every erroneous slice, we...