Desing And Implementation Of Motion Vector Predictor In H.264Video Decoder

With the rapid development of integrated circuit technology, people graduallyincreasing demand for quality in high-definition television and videocommunications. In the background of matures technologies and the market’seagerly calling, H.264protocol came into being. Although the encoder and decoderwhich in line with the H.264protocol have an excellent compression ratio, but thepresence of high computational complexity is a big problem, which makes the codecdesign and verification more and more difficulty because of the increasing of newfunctions and features with the new updates of the protocol.In this paper we take the first step to analysis and demonstrate the importanceof motion vector prediction process of the H.264video decoder. With the producingand conversion in the life cycle of the data stream, Introduction principle of eachmodule of the encoder and decoder and the inter prediction process in decoder.The two difficult in design algorithms and stored were analyzed and optimizedseparately. In terms of algorithms according to the workflow of the MV predictiondecoder, with the functional unit of the basic module proposed the overall structuremodule. Ports and timing were designed on this basis. Positioning and effectivenessof adjacent modules to be judged on the basis of the agreement, the value whichcomes from direct prediction or median operation was pulsed with the residualstransformed from the bit stream, and then MV is prepared.In terms of storage, through in-depth analysis of the decoding process, labelingalgorithm and optimization mechanism was proposed. Thereby significantlyreducing the occupied storage resources, on the other hand appropriate strategies forthe actual demand for high-definition video decode was propose. Finally,considering all the analysis above the detailed sub-module design come up.In the verification part, by learn and analysis of strategies and methods ofverification. Considering of the actual situation we selected gray box testing,dynamic simulation software functional verification and FPGA-based verificationmethods. The designed prediction module was integrated into the open sourcedecoder, optimized decoder, takes function test, build SoC framework, and testedthe IP on FPGA, After compare to the results of the test data which made by JMmodeling software, to make sure whether this MV module is correct.