Efficient implementation of integrall image algorithm on NVIDIA CUDA
Efficient implementation of integrall image algorithm on NVIDIA CUDA
复制标题
积分图像算法在NVIDIA CUDA上的高效实现
DOI：
10.1109/aset.2018.8379824
复制
发表时间：
2018
期刊：
影响因子：
--
通讯作者：
Mohamed Atri
中科院分区：
文献类型：
--
作者：
Mouna Afif;Yahia Said;Mohamed Atri
GPU计算通过尝试优化因其特定功能而需要大规模并行计算的任务，使执行更高效的实施结果成为可能。这是实现的GPU算法数量增加的主要原因。与CPU相比，GPU计算在加速算法处理方面证明了其有效性。本文介绍了用CUDA编程语言在GPU上实现积分图像算法的方法。在许多图像处理算法中，积分图像是一个重要而关键的步骤。我们还比较了我们的算法在CPU和GPU上的性能以及获得的加速。我们还将我们的算法与其他使用编程语言CUDA的GPU实现进行了比较。实验结果表明了该算法的有效性。与其他CPU和GPU实现相比，我们获得了更高的加速比。
GPU computing makes it possible to perform more efficient implementation results by trying to optimize tasks that require massively parallel computing due to its particular capabilities. This is the main reason for the increase in the number of implemented GPU algorithms. Compared to the CPU, GPU computing has proved its efficiency in accelerating the processing of algorithms. This paper presents an implementation of the integral image algorithm on GPU by using the programming language CUDA. Integral image is important and crucial step in many image-processing algorithms. We also show a comparison between the performance of our algorithm on CPU and GPU on as well as the accelerations obtained. We also compare our algorithm with other GPU implementations using the programming language CUDA. The achieved results show clearly the efficiency of our algorithm. We achieve high speedup results comparing to other CPU and the GPU implementations.