多图像同态滤波的 CPU 和 GPU 并行计算磁

摘要

采取 CPU 分发图像滤波任务和回收滤波结果、将多个图像数据划分分配给多个 GPU 及其线程块、GPU 调用核函数库对图像进行傅里叶变换和反傅里叶变换的方法，设计实现了 CPU 和 GPU 协同计算的多图像同态滤波并行算法。实验结果表明，给出的多图像同态滤波并行算法高效，与单 GPU 计算的并行算法相比，多 GPU 协同计算的并行算法显著缩短了多个图像同态滤波处理所需的时间。%The multi-image homomorphic filtering task is distributed and the filtering results is collected by CPU ,the given multiple images are assigned to the multiple GPUs and their thread blocks ,GPU core functions are called to execute Fourier transform and inverse Fourier transform ,and then the multi-image homomorphic filtering algorithm using CPU and GPUs cooperatively parallel computing are designed and implemented .The experimental results show that the given parallel filtering algorithms are efficient ,and compared with the parallel algorithm using single GPU computation ,the one using mul-tiple GPUs computation can remarkably reduce the running time .