Research on Orthorectification of Remote Sensing Images Using GPU-CPU Cooperative Processing

This paper first introduces the basic processing flow of orthorectification of remote sensing images and analyzes its parallelism, then proposes a fast GPU-CPU cooperative processing algorithm based on CUDA, which realizes fine-grain parallel image resampling in single GPU and coarse-grain parallel processing between multi GPUs. Three speedup strategies are adopted to improve the implement performance based on parallel architecture and hardware characteristic of GPU. Finally a series of orthorectification experiments using PSC-2N desk supercomputer and UCD aerial digital images are implemented to prove the advantages of GPU-CPU cooperative parallel computing.