基于FPGA的AVS编码器帧内预测实现

Intra Frame Prediction for AVS Encoder based on FPGA

针对帧内预测的快速算法,由于DSP架构软件顺序执行的局限性难以满足实时性要求,而FPGA以其高速的计算速度和强大的并行处理能力成为H.264和AVS编解码的理想平台。本文在FPGA平台上采用资源共享、高并行和多流水线结构实现了亮度帧内预测算法。该方法在分析AVS帧内亮度5种预测模式的基础上,将像素预测与模式判决在一个模块中完成,并且利用各模式预测的相似性,实现运算单元共享和多种模式并行执行,兼顾了处理速度和实现代价。仿真及综合结果表明该设计能够完全满足标清(704×576,30f/s)数字视频的实时处理要求。

The fast algorithm of intra prediction is difficult to meet the real-time demand on DSP because of the limitations of se-quential order of DSP software architecture;while FPGA with the high calculation speed and strong parallel processing ability has become the ideal platform for H.264 and AVS codec.In this paper,a method which applies the resource-sharing,high parallel and pipeline structure has been put foward.The method,in the analysis of 5 prediction modes of AVS intra luma and their similarity in operation,has put the pixel prediction and mode decision in a module and realized the sharing of public operation unit and concurrent execution of a variety of modes on consideration of both processing speed and realizing costs.The simulation and comprehensive results show that the design has been completely meet realtime processing requirements of the standard（704 x 576,30 f/s）digital video.