Research on CUDA-Based SIFT Registration of SAR Image

Abstract

This paper proposes an implementation and optimization of SIFT algorithm for SAR (Synthetic Aperture Radar) images. It improves the SIFT algorithm efficiency using the graphics processing unit (GPU) architecture based on Compute Unified Device Architecture (CUDA) framework, to attend the request of real-time remote sensing image processing. The experiments of large-size SAR image demonstrate that the algorithms achieve 57.8 times speedup.