Research on Grid-Enabled Parallel Strategies of Automatic Wavelet-based Registration of Remote-Sensing Images and Its Application in ChinaGrid

Abstract

ChinaGrid is an important project aiming to providing high performance services in a grid computing environment.One of the applications offered by ChinaGrid, image registration is described. It is an important processing step in many applications of remote sensing area, which is requires intensive computing power. Firstly, the serial and existing parallel strategies of wavelet-based automatic image registration are overviewed. And then, some grid- enabled optimization for two of them, Hybrid-Parallel (HP) and Group-Parallel (GP), is given. Thirdly, all of these parallel strategies are reevaluated for grid environment theoretically and a performance testing of these parallel algorithms is done on a simulated grid platform. Experimental results show that our optimization is effective but the basic simple parallel strategy Parameter-Parallel (PP) is the best choice for grid. The parallel algorithms proposed here have been integrated into related service system of ChinaGrid.