采用GPU加速的压缩感知图像恢复算法
Accelerate Compression Sensing Reconstruction Algorithms Using GPU
-
摘要: 压缩感知(Compressed Sensing, CS)的信号重构部分需要进行大数据量的计算, 然而传统的CPU对大量的矢量计算并没有优势.为了解决这一问题, 我们以CUDA作为并行计算架构, 通过GPU-CPU并行编程技术, 实现了三种快速高效的压缩感知图像恢复算法, 包括正交匹配追踪OMP算法、两步阈值迭代TwIST算法和线性Bregman算法.Abstract: The signal reconstruction in compressive sensing (CS) requires a large amount of data processing. However the traditional CPU has no advantage on vector calculation. To deal with this issue, we design a parallel computing architecture using CUDA. Based on GPU-CPU parallel programming technology, and we realize three fast and efficient CS image reconstruction algorithms, including OMP, TwIST, and linear Bregman algorithm.