图像处理的GPU加速技术研究  

Research of GPU acceleration techniques for image processing

作　　者：詹洪陈[1] 袁杰[2] 

机构地区：[1]南京大学金陵学院,江苏南京210089 [2]南京大学电子科学与工程学院,江苏南京210093

出　　处：《现代电子技术》2012年第20期87-90,94,共5页Modern Electronics Technique

基　　金：国家自然科学基金资助项目(60501003)

摘　　要：通过Matlab和Visual C++两个平台,实现了对图像工程的并行加速处理,并且通过Jacket,CUDA两种加速方案的介绍,进一步了解使用GPU高性能并行计算的工作流程以及性能效益。最后,给出了通过并行处理之后的关于两个图像工程计算性能的测试结果及比对。结果证明,经并行处理后的图像工程在计算效率方面有显著提高,结果精确,计算耗时小。The parallel accelerated processing of the image project was achieved through Matlab and Visual C＋＋.By introducing the acceleration schemes of Jacket and CUDA,the working process of GPU＇s high-performance parallel computing technology and its performance benefits were got.give The test and comparison results of the computing performance of the two image projects which have got through parallel processing are given.The results show that parallel processing technology dose improve the computing efficiency of image projects with accurate data and less time consumption.

关 键 词：并行处理技术 JACKET CUDA MATLAB 

分 类 号：TN919.34[电子电信—通信与信息系统] TP751[电子电信—信息与通信工程]