计算机应用与软件2012，Vol.29Issue(9)：18-22,5.DOI:10.3969/j.issn.1000-386x.2012.09.005
基于特性分析的SURF加速方法研究
RESEARCH ON SPEED-UP METHOD OF SURF THROUGH PROPERTIES ANALYSIS
摘要
Abstract
To resolve the problem of a little bit low speed the typical image feature extraction algorithm SURF (speeded up robust feature) has, in the paper we do an in-depth analysis on the algorithm properties in regard to its running time, instruction type, memory access, hot spots,and etc. On this basis, we optimise the SURF algorithm by reconstructing the judgment logic, the memory accessing and data organisations,the library functions implementation and other strategies. Moreover,we analyse the role of speedup of each optimisation strategy on the algorithm through experiments. And experimental data show that these optimisation strategies can effectively improve the running efficiency of SURF in the premise of guaranteeing 100% accuracy of the algorithm, and 2. 75X ultimate performance improvement is achieved.关键词
SURF(speeded up robust feature)/特性分析/加速Key words
SURF/ Properties analysis/ Speedup分类
信息技术与安全科学引用本文复制引用
胡益斌,肖菁,戴卓方,李戈,张为华..基于特性分析的SURF加速方法研究[J].计算机应用与软件,2012,29(9):18-22,5.基金项目
国家自然科学基金项目(60903015) （60903015）
Intel大学合作项目 （）
国家高技术研究发展计划(2009AA012201) （2009AA012201）
中国科学院计算技术研究所体系结构重点实验室开放课题(ICT-ARCH2009082009). （ICT-ARCH2009082009）
