CUDA架构下的灰度图像匹配并行算法
Parallel Grey-Level Image Matching Algorithm with CUDA
-
摘要: 提出一种NVIDIA CUDA架构下的灰度图像匹配算法,利用GPU加速灰度图像的匹配过程。实际的测试结果表明,在现有实验环境中,对同一图像,在不损失匹配精度的前提下,在GPU上使用CUDA实现的灰度图像匹配并行算法比在CPU上使用MPI实现的灰度图像匹配并行算法快了40多倍,性能得到了显著提高,从而使灰度图像匹配应用于如交互式系统等实时应用成为可能。Abstract: This paper proposes a grey scale image matching algorithm with NVIDIA CUDA in order to accelerate the process using GPU. Our test shows that when matching on the same image, the CUDA program of grey-level image matching algorithm is 40 times faster than the CPU with MPI version, without the decline of matching precision. The result shows that the grey scale image mathing can be applied to real-time applications such as interactive systems.