摘要
GPU通用计算的发展为图像处理提供了新的技术手段,目前一些图像处理算法已经移植到GPU中且得到了较好的加速效果,但是并没有形成通用的并行设计模型,推广性差。为此,论文在研究GPU结构特点以及已有的GPU图像处理并行算法的基础上,提出了单GPU和多GPU图像并行处理设计模型,并通过图像平滑并行算法在GPU中的执行验证了该模型的有效性。试验结果表明,该模型在GPU图像处理中通用性较好,为更多算法的并行化处理提供了理论借鉴。
The development of GPU general computing provided a new technique for image processing, and some image processing algorithms had been transplanted to GPU and got a better acceleration effect. But it could not form a general parallel design model. Thus, based on the study of the characteristics of GPU structure and the existing parallel algorithms of GPU image processing, this paper proposed a parallel image processing design model of single GPU and multi GPU. And this paper took the algorithm design of parallel image smooth as an example to testify the effectiveness of this model. The experimental results showed that the proposed model had good generality, and it could provide a theoretical reference for the parallel processing of more algorithms in the GPU image processing.
出处
《舰船电子工程》
2016年第11期74-76,134,共4页
Ship Electronic Engineering
关键词
图像算法
并行化
GPU
模型
通用性
image algorithms, paral lel, GPU, model, generality
