基于流体系结构的帧内预测算法优化设计

为高效实现H.264多模式帧内预测，解决其计算复杂度高造成的计算压力，本文根据H.264帧内预测算法的计算密集与数据并行的特征，基于流处理执行模型提出适用于Imagine流体系结构的分组帧内预测流算法，并采用长流分段技术进行优化设计.实验结果表明，H.264帧内编码器流实现对1280×720高清视频编码帧率达45.9fps，满足实时性需求.

Optimized Design of Intra Prediction Algorithm on Stream Architecture

Multi-mode intra prediction in H.264 is computationally expensive.For efficient implementation to reduce computing pressure,according to computation intensity and data parallelism in intra prediction,a multi-group intra prediction stream algorithm suited for imagine stream architecture is proposed based on stream execution model.In addition,strip mining technology is applied into streaming implementation to design an optimized strip-mined multi-group intra prediction stream algorithm.Experimental results show that H.264 intra coder using optimized stream algorithm processes 1280×720 HD video at 45.9fps,satisfying the requirements of real-time video applications.