摘要
设计了一种针对图像处理类应用程序的并行性分类的方法,并使用工具初步实现了对程序的分类.还研究了图像处理算法的并行性.通过对图像处理算法的主要特征进行分析,得出图像处理算法包含多种并行特征的结论,分别从图像处理程序的混合结构的模型,以及程序中的依赖关系和并行特征两方面进行了分析.结合对程序的分类方式,使用LLVM编译架构初步实现了自动对程序中的程序段进行识别其中的分类模式.
According to the main characteristics of the image processing algorithm,the conclusion that the image processing algorithm contains a variety of parallel features is analyzed from the model of the mixed structure of the image processing program and the dependencies and parallel features in the program.Then,according to the analysis of the dependencies in the loop,the parallelization of the different levels of the program is found,and the loops are classified.The different categories correspond to the different execution units,and the typical image processing procedures are further analyzed.
引文
[1]代媛.基于GPU并行计算的图像压缩技术研究[D].杨凌:西北农林科技大学,2014.
[2]Silva B D,An B,D'Hollander E H,et al.Comparing and combining GPU and FPGA accelerators in an image processing context[C]∥International Conference on Field Programmable Logic and Applications.IEEE,2013:1-4.
[3]张玮雄,刘建霞.数字图像处理技术的发展现状及趋势[J].科学之友,2012,33(6):153.
[4]龚志远.基于图像分析技术的轴承故障诊断[J].煤矿机械,2005,26(11):171-173.
[5]郭启军,潘炼,钱新恩.基于图象识别的高抛光透射面缺陷检测[J].计算技术与自动化,2008,27(4):95-97.
[6]许琳娟,曹文洪,刘春晶,等.图像处理技术在推移质运动颗粒参数提取中的应用[J].长江科学院学报,2017,34(1):1-5.
[7]Gonzalez R C,Woods R E.冈萨雷斯数字图像处理[M].北京,电子工业出版社,2003.