Research On C Modeling Of AVS Video Decoder

As the second generation coding standard, AVS (Audio and Video coding Standard) is our own independent Intellectual Property. The AVS standard includes four main technical part- system, audio, video and media copyright management, and a support standard-consistency test. It provides a complete solution for digital audio and video industry. AVS video is the second part of AVS standard. Now, the baseline-profile of video has been established and is already on the way of industrialization. And the X-profile is also on the stage of researching and discussing.Firstly, the thesis mainly discusses the main technical modules in the avs video standard. Secondly, according to the configuration of common DSP, a C model is constructed based on the reference video decoder. The core algorithm of avs video decoder are presented and studied in detail. Then, with the research on the reference implementation of avs video decoder, the decoding flow graph is drawn. After farmillar with the code of the decoder and the difference between the reference decoder and the C-model, the C-model construction of the video decoder is on the way. The specific steps include: the adjustment of the decoder structure; the package of some function and the reasonable compression of the memory space of constant tables in the decoder; the removal of some functions related to the special systems; the adjustment of all kinds of buffers in the process of decoding in order to compress the on-chip SRAM space; the split of the huge structure in the decoder; the on-chip and off-chip partition of the memory space of motion-vector, the reference index and the reference frame data. Besides, the mechanism of the deblock filter is fixed so as to reduce the interactivity between the on-chip SRAM and off-chip DRAM, and the mechanism of error report is added in the decoder, etc.Then, the C-model is profiled. A summarization and analysis of the C-model is made from aspects of operation-complexity and memory space complexity.At last, ConformanceTest is made on the C-model, and the result proves that the C-model decoder can process the input coded video stream in the right way.