摘要
傅里叶变换是地理栅格数据处理的重要算法之一。但当面对超大型栅格数据时,传统的单个进程方法无法快速地处理出结果;而且由于多数方法需一次加载所有数据,所以很多程序难以处理超过系统内存大小的数据。该文提出一种面向超大型栅格数据的并行快速傅里叶变换算法,通过"列暂存文件"作为交换空间,实现整个数据的分步骤处理,在有限的内存空间下,可以并行地分步骤、分块地处理整个数据。实验表明,该算法可以显著提高傅里叶变换运行速度,并一直占用较低的内存空间。
Fourier transform is an important algorithm in the field of geographic data processing. However, when process extremely large raster data, the traditional single-process algorithm cannot product results quickly; and most method can＇t compatible with the data size more than the system memory size because they need load all the data in to memory. This paper presents a novel extremely large raster data parallelized fast Fourier transform algorithm,through ＂temporary column files＂ as exchange space, it can parallelize and partitioned process the entire raster data step-by-step with limited memory space. Experiments show, the proposed method can significantly improve the Fourier transform running speed,and only need relatively limited memory space when the program executing.
出处
《地理与地理信息科学》
CSCD
北大核心
2013年第4期52-55,98,共5页
Geography and Geo-Information Science
基金
国家863计划项目子课题(2011AA120302)
国家自然科学基金青年基金项目(41101384)
关键词
栅格
并行
傅里叶分析
超大型数据
MPI
raster data
parallelize
Fourier transform
extremely large data
MPI
