| 引用本文: | 文梅,任巨,伍楠,等.流化H.264编码的数据并行方法.[J].国防科技大学学报,2011,33(3):66-71.[点击复制] |
| WEN Mei,REN Ju,WU Nan,et al.Data Level Parallel Method of Streaming H.264 Coding[J].Journal of National University of Defense Technology,2011,33(3):66-71[点击复制] |
|
| |
|
|
| 本文已被：浏览 8175次 下载 5937次 |
| 流化H.264编码的数据并行方法 |
| 文梅, 任巨, 伍楠, 苏华友, 荀长庆, 张春元 |
|
(国防科技大学 计算机学院，湖南 长沙 410073)
|
| 摘要: |
| 流模型是一种源于媒体处理的新型并行计算模型。然而H.264编码作为媒体处理领域中的重要应用，在与流模型适配时，却出现重要障碍，主要是由于相关性问题极大地限制了流化过程中数据级并行的开发。针对这个问题，首先根据编码算法分析了编码过程中遇到的相关性问题，然后为各主要编码模块分别提出解除相关性限制的数据并行编码方法。采用这些方法能够保证大数据并行粒度，最终达到高效并行流化编码的目的。实验表明采用数据并行流化编码方法，帧间预测、帧内预测、熵编码和去块滤波模块都获得了显著加速。 |
| 关键词: H.264编码 流模型 并行计算 数据并行 |
| DOI： |
| 投稿日期：2010-07-30 |
| 基金项目:国家自然科学基金资助项目(60703073，60903041，61033008) |
|
| Data Level Parallel Method of Streaming H.264 Coding |
| WEN Mei, REN Ju, WU Nan, SU Huayou, XUN Changqing, ZHANG Chunyuan |
|
(College of Computer, National Univ. of Defense Technology, Changsha 410073, China)
|
| Abstract: |
| Stream processing model is a newly emerging parallel processing model, which is derived from media processing. However, as an important application of media processing, H.264 coding encounters problems during its streaming procedure. The main reason is that dependency limits the data level parallel exploitations. To solve this problem, data level parallel methods for different coding modules were developed by the analysis of coding algorithm. By adopting the method, key coding modules of H.264 can attain large data parallel granularity. Experiment results show that the data level parallel streaming inter/intra prediction, CAVLC, and the deblock filter can achieve outstanding acceleration. |
| Keywords: H.264 coding stream processing model parallel processing data level parallel |
|
|
|
|
|
