摘要
研究了如何利用GPU来加速视频解码,概述了MPEG-2视频解码的系统框架,论述了MPEG-2视频解码在Linux下以XvMC(X video motion compensation)为API并基于通用可编程GPU的实现过程,重点讨论了MPEG-2视频解码中IDCT(inverse discrete cosine transform)和运动补偿的实现,提出了新的优化算法。MPEG-2视频解码算法具有一定的通用性,实验结果表明,与传统的解码方式相比,该解码器不仅能加速视频解码,还能有效降低CPU的利用率和电脑的功耗。
The study on leveraging the GPU to accelerate the video decoding is presented,the system framework of decoding of MPEG-2（moving pictures experts group） is summarized,the realization of MPEG-2 hardware decoding using XvMC（X video motion compensation） based on general programmable GPU is expounded,and especially the implementation and optimization of motion compensation and IDCT are discussed.The algorithm of MPEG-2 decoding can be used in many situations.The experimental results show that significant speed-up can be achieved,CPU utilization and compute energy can be reduced significantly by utilizing the GPU power.
出处
《计算机工程与设计》
CSCD
北大核心
2010年第10期2253-2256,共4页
Computer Engineering and Design
基金
上海市重点学科建设基金项目(J50103)
