An asynchronous parallelized and scalable image resampling algorithm with parallel I/O

Abstract

Image resampling which is frequently used in remote sensing processing procedure is a time-consuming task. Parallel computing is an effective way to speed up. However, recent parallel image resampling algorithms with massive time-consuming global processes like I/O, always lead to low efficiency and non-linear speedup ratio, especially when the amount of computing nodes increases to a certain extent. And what's more, the various geo-referencing related to different processing applications caused a real problem of code reuse. To solve these problems, Parallel Image Resampling Algorithm with Parallel I/O (PIRA-PIO) which is an asynchronous parallelized image resampling algorithm with parallel I/O is proposed in this paper. Parallel I/O of parallel file system and asynchronous parallelization which using I/O hidden policy to sufficiently overlap the computing time with I/O time are used in PIRA-PIO for performance enhancement. In addition, the design of reusable code like design pattern will be used for the improving of flexibility in different remote sensing image processing applications. Through experimental and comparative analysis, its outstanding parallel efficiency and perfect linear speedup is showed in this paper. © 2009 Springer Berlin Heidelberg.