Deeply Pipelined DSP Solution to Key Modules in H. 264
With the rapid develpment of microprocessor,embedded multimedia products are gradually becoming the mainstream in the market.However,the high coding efficiency enabled by the H.264 video compression standard comes with substantially greater algorithmic complexity as compared to that of existing standards.And this additional complexity results in many difficulties in the implementation and optimization tatsks.This paper analyzes the algorithms of the two time-consuming modules of integer transform and motion estimation in H.264.Then optimizes the two modules based on the extended instruction set of C64x/C64x+.Finally,deeply pipelined DSP solutions to two modules are presented in this paper.The experiment results show that optimizing parallel assembly can make the codes moreefficient.
H.264.DM647JCTInteger Cosine Transform motion estimation pipeline optimizing parallel assembly
Jinxiu Zhu Ning Cao Yushan Chen Guoxuan Li
College of Computer and Information Engineering Hohai Univ.Changzhou 213022 China College of Electrical Engineering Hohai Univ Nanjing 210098
国际会议
The Fifth IEEE International Symposium on Embedded Computing(SEC 2008)(第五届IEEE国际嵌入式系统计算会议)
北京
英文
107-112
2008-10-06（万方平台首次上网日期，不代表论文的发表时间）