A MPI-based parallel pyramid building algorithm for large-scale RS image

Building pyramid for remote sensing (RS) image is an effective way to achieve image multi-resolution organization, and also an important way to improve performance of image browsing. For large-scale remote sensing image, traditional sequential pyramid building processing is a time consuming task in many applications. By taking advantage of multi-core, multi-node cluster computing environment and parallel processing mechanism, a MPI-based parallel algorithm is proposed, which can greatly improve the performance of pyramid building. The algorithm has a good scalability and can easily be extended to a considerable number of nodes. Experimental results show that the proposed algorithm has better acceleration effect compared to the sequential methods, and there is a positive correlation between the acceleration effect and image size. For large remote sensing image (in our case 46 GB), the parallel algorithm can be about 10 times faster than GDAL.