Moving Picture Algorithm Implementation Of XAVS Encoder And Overall Performance Optimization

AVS is second-generation source coding standards with independent intellectual property rights in China. It has three characteristics: advanced, independent, open. At present, in terms of the realization of the software, the Working Group's reference software is only available, but the reference software has shortcomings in both program structure and module realization, which is not conducive to the promotion of AVS .Therefore, developing a high performance, structured AVS encoder is most urgent and important. In this background,this dissertation responded to the call of AVS Working Group to develop the open source encoder xAVS.After carefully studied the AVS1-P2 standard (AVS-video),this dissertation developed the high performance encoder xAVS, and mainly completed the following work:(1) Developed the objective task of xAVS encoderThe objective task for xAVS is to develop high performance encoder based on x86 platform to meet the AVS-video standards, to support the compression of multimedia sources, to support rate control and rate distortion optimization, and to package as a standard transport stream.(2) The Implementation of moving picture algorithmThis dissertation designed the inter-frame coding mode in which mode selection and coding used the macroblock as a unit for processing while interpolation and filtering was to be processed by frame after the completion of the whole macroblock coding. In the mode selection, it used characteristics of Skip mode, adopted the method of decision-made beforehand in Skip mode. This method eliminated a large number of time-consuming operation, and improved the encoding speed. Being tested, xAVS had better performance than the reference software rm52j.(3) Scene cut detectionConsidering a significant decline in the quality of coding when the scene cut in P frames, this dissertation proposed scene cut detection algorithm for P frames. Via coding default parameters it calculated the upper and lower thresholds and used the thresholds to attain the scaling factor of the intra prediction cost. Comparing the cost of the current inter prediction block and the counterpart of intra prediction block to detect scene cut, when the former is greater than the latter, then re-use I frame as a scene cut frame to code, and vice versa.(4) Packaging for transport streamIt analyzed structures of the AVS-video stream and the transport stream. By parsing the sequence start code and picture start code of AVS stream, PES header information is added before these two start codes. According to 188B packet data structure of TS stream, it filled the corresponding additional information in the PES group, and packaged for the transport stream. Packaged transport stream can be played in storm video. (3.09.12.16 version or later)Furthermore, for the transplantation to DSP, this dissertation has separated I, P, B frames. It extracted image type selection from macroblock layer to image layer, removed the irrelative frame code, so as to enhance the hit rates of CACHE.In summary, this dissertation researched and developed the xAVS encoder. xAVS encoding performance is better than reference software rm52j and not less than x264, tested by National Engineering Laboratory For Video Technology. Through identification of scientific and technological achievement, organized by The ShanXi Science and Technology Department, xAVS encoder reached the international advanced level. xAVS team was highly prairied by AVS group and achieved the AVS special contribution award of 2009.