基于CUDA的高速并行高斯滤波算法  

CUDA-based algorithm for high-speed parallel Gaussian filtering

作　　者：卢文龙[1] 王建军[1] 刘晓军[1] 

机构地区：[1]华中科技大学机械科学与工程学院,湖北武汉430074

出　　处：《华中科技大学学报（自然科学版）》2011年第5期10-13,共4页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家自然科学基金资助项目(51075169);湖北省自然科学基金资助项目(2008CDA029)

摘　　要：为加快表面三维形貌分析中高斯滤波算法的执行速度,提出了一种基于计算统一设备构架(CUDA)的高斯滤波算法来实现高速并行处理.分析高斯滤波算法原理和CUDA并行计算体系,将CUDA并行计算技术引入到表面分析领域.针对高斯滤波数据间依赖性弱和CUDA采用单指令多线程(SIMT)执行模型的特点,总结出适合于CUDA的并行高斯滤波算法流程.实验证明:该方法与CPU串行处理方法相比,其加速比达到40倍以上,可以有效提高数据处理能力.In order to speed up the Gaussian filtering in three-dimensional surface texture analysis, an efficient method based on compute unified device architecture （CUDA） implemented on graphic pro- cessing unit （GPU） was designed. CUDA parallel computing technology was introduced into the sur- face texture analysis area by analyzing principle of Gaussian filtering and CUDA computing architec- ture. Parallel Gaussian filtering algorithm used in CUDA was given for the characteristics of low de- pendency of Gaussian filtering and SIMT （single instruction multiple thread） execution model of CUDA. Experiments prove that the calculation speed based on CUDA is 40 times faster than that of traditional sequence algorithm based on CPU, can effectively improve data processing capability.

关 键 词：高斯滤波 计算统一设备构架 图像处理器 表面形貌 并行处理 

分 类 号：TH132.1[机械工程—机械制造及自动化]