摘要
为了提高新一代音视频编解码技术标准AVS的编码速度,利用OpenMP在多核处理器平台上研究并实现了AVS的GOP级、条带级、帧级和基于任务队列模型的帧级并行编码算法。对CIF格式的视频序列进行了测试,在四核处理器平台上加速比最高能达到3.82x。另外,基于任务队列模型的帧级并行算法在保持图像质量不变的基础上解决了帧级并行算法加速比偏低的缺点。实验结果表明,OpenMP是一种简单而有效的并行化编程工具,基于OpenMP的各个AVS并行编码算法与原串行算法相比,编码速度都有显著提高。
To increase the encoding speed of AVS,the parallel encoding algorithms including GOP-level,slice-level,simple frame-level and frame-level based on task queuing model are researched and realized,based on OpenMP and multi-core processors.CIF format sequences are tested and speedup can obtain 3.82 on a four-core processor system.In addition,the algorithm of frame-level based on task queuing model solves the problem of low speedup in simple frame-level algorithm and maintain the same image quality.The experimental results prove that OpenMP is a simple and effective tool for parallel programming and these parallel algorithms for AVS based on OpenMP can improve encoding speed signifiantly compared with original serial algorithm.
出处
《计算机工程与设计》
CSCD
北大核心
2010年第10期2257-2261,共5页
Computer Engineering and Design
作者简介
作者简介：胡文安（1981-），男，硕士研究生，研究方向为视频编解码技术；
于鸿洋（1964-），男，副教授，研究方向为多媒体通信、数字音视频处理。E-mail：anhoho@qq．com
