Research On Compression Of Video Signal Based On Wavelet Transform

Video signal is an important medium of human perception of nature,and it plays an important role in the field of multimedia communication and signal processing.However,video signals need more data representation than other media since video signals contain both static and dynamic information,which means more storage space and transmission bandwidth are needed.In order to solve the problem of large amount of video data,video coding technology arises at the historic moment.The purpose of video coding is to reduce the bit rate of video signals while the video quality is kept unchanged.As an excellent time-frequency analysis tool,wavelet transform can perform multi-scale analysis of signals and better represent the local time-frequency characteristics of signals.As a result,wavelet transform has been widely applied and developed in academia and industry since its emerge.The principle of codec and bitstream extractor in the T+2D wavelet video encoding scheme is systematically studied and the advantages and disadvantages on wavelet video encoding as scalable video encoding are discussed.The biggest advantage of wavelet video coding is that the encoder can generate the embedded bit stream with frame rate,resolution and quality scalable as the virtue of the multi-resolution property of wavelet transform.The wavelet video coding based on motion compensated temporal filtering adopts the open loop structure,and it avoids the decoding process in the coding successfully,reducing the burden of the encoder to a certain extent,but the high complexity of the wavelet video coding still makes it difficult to use in real-time applications.Therefore,in order to solve the problem that the wavelet video codec based on the T+2D coding scheme is too complex,the complexity of wavelet video encoder is reduced from two aspects: algorithm optimization and platform optimization.(1)motion estimation algorithm is improved.Motion estimation is the most complex module in the T+2D wavelet video encoding.Thus it is necessary to reduce the complexity of motion estimation algorithm without the bit rate varied obviously.The TZSearch algorithm is first applied to the wavelet video encoding.Combined with the characteristics of block matching in the wavelet video and the relevance of motion search process,an improved TZSearch algorithm is proposed from the aspects of static block detection,stop search criterion and adaptive search range.In order to verify the performance of the algorithm,the video sequences with different resolutions and different motion complexity of textures are tested.The results show that the time complexity is reduced by the improved motion estimation algorithm compared with the original algorithm,while the bit rate and video quality are basically unchanged.(2)Optimal porting of the wavelet video codec on DSPs.The fixed-point DSP chip TMS320DM642 is chosen as the experimental platform.In order to speed up the encoding and decoding speed based on the platform,the fixed-point algorithm of the wavelet video encoding is implemented,and then the necessary simplification of the wavelet video encoding module is carried out,optimized by using software pipelining,intrinsics,eliminating the correlation of memory,linear assembly optimization etc.In these ways the parallel instructions and hardware resources are taken full use of,improving the execution efficiency of programs.Experimental results show that the number of clock cycles consumed by the simplified and optimized wavelet video coding program is significantly decreased by at least 90%,making it possible to be a real-time codec on the hardware platform.