An Optimized Image Mosaic Algorithm with Parallel IO and Dynamic Grouped Parallel Strategy Based on Minimal Spanning Tree

Image mosaic on a great number of remote sensing images with large scale is commonly I/O intensive and time-consuming. However, parallel computing is recently considered as an effective way to solve the problems of enormous computation caused by its complicated algorithm and the massive data amount. But traditional parallelization of mosaic algorithm doesn´t optimize the serial processing pass of mosaic algorithm according to the parallel strategy. Furthermore, the registration and blending in mosaic is time-consuming, but there is no effective parallel strategy of data splitting to solve this problem. And what´s more, for massive remote sensing data the problem result from the I/O performance of the mosaic algorithm will bring perform bottle. To solve these problems, An Optimized Image Mosaic Algorithm with Parallel I/O And Dynamic Grouped Parallel Strategy Based On Minimal Spanning Tree is proposed in this paper. The processing procedure of the traditional algorithm is optimized to meet the need of the parallel strategy. What´s more, effective parallel strategy of data splitting is adopted in the time-consuming part registration and blending. In addition, the multi-thread parallel I/O strategy which is overlapping I/O and computing time is adopted to speed up the algorithm efficiency. Through experimental and comparative analysis, its outstanding parallel efficiency and perfect linear speedup is showed in this paper.