A GPU-based Parallel Implementation of Compressive Sampling Reconstruction for SAR Image Compression

Compressive Sampling (CS) technique has been adopted to compress Synthetic Aperture Radar (SAR) images. However, due to the mass data of high resolution SAR images, the reconstruction of compressive sampling generates huge computational load, making it impossible to run on traditional CPU at real time. To solve this problem, this paper attempts to implement the reconstruction produce in parallel based on Graphics Processing Unit (GPU) device. The results show that the GPU-based implementation is faster (up to 8.8 times) than the CPU-based implementation.