A mapreduce approach for processing large-scale remote sensing images

Abstract

This paper introduces a practical approach to process large-scale remote sensing images based on MapReduce program model. The proposed approach breaks the global large-scale remote sensing images processing into a set of different resolutions localized domains (tiles). Within each tile, the different remote sensing image processing algorithm can be applied. The critical step in this approach is how to integrate the MapReduce model smoothly with the large-scale remote sensing images processing. The proposed approach adopts the idea of tile pyramid which partition the global into multi-resolutions tiles. With the tile pyramid model, each remote sensing image will be partitioned and assigned to multi-resolutions tiles that it overlapped, and then all images assigned to the same tile will be assembled to form the final tile image; finally, the per-tile image processing can be performed. The experiment results show the feasibility, efficiency and scalability of solving large-scale remote sensing images problems with the proposed approach.