摘要
为了实现数字全息中粒子三维位移场的快速重建,利用CUDA架构将原有串行的重建步骤进行二级并行化处理。同时,结合多线程技术实现线程间的并行处理,以进一步提高粒子三维位移场重建速度。模拟和实验结果表明:在不同重建平面数量和不同分辨率情况下,采用GPU完成平面重建的运算速度至少达到CPU运算速度的166倍。将其实际应用获得全息图序列以进行三维位移场的并行运算,其处理的单幅全息图为512像素×512像素,并行程序完成512个平面重建及其相关后处理所需时间为268 ms,即3.7帧/s,其运算速度为采用CPU进行串行计算的40.6倍。
In order to realize the fast reconstruction of the 3D displacement field of particles,a new method using the compute unified device architecture to perform the two level parallel processing instead of original serial reconstruction steps was proposed. At the same time, multi-threading technology was applied to achieve parallel processing for the sake of improving the reconstruction speed of 3D displacement field further. Simulation and experiment results showed that the operation speed of reconstruction images using graphics processing unit was at least 166 times as fast as CPU with different reconstruction plane number and different resolution. It was applied to the actual sequence of holograms to obtain parallel computing of 3D displacement field,the single hologram was 512 pixels × 512 pixels. The total time of reconstructing 512 images and related post processing was268 milliseconds. In other words,processing speed of holograms was 3. 7 frames per second,and it was 40. 6 times as fast as CPU which processed the data using serial mode.
作者
朱岗
杨岩
ZHU Gang;YANG Yan(School of Mechanical Engineering,Chongqing University of Technology, Chongqing 400054, Chin)
出处
《重庆理工大学学报（自然科学）》
CAS
北大核心
2018年第4期111-118,共8页
Journal of Chongqing University of Technology：Natural Science
基金
国家自然科学基金资助项目(11272368)
重庆市教委科学技术研究项目(KJ1600929)
重庆理工大学青年科研项目星火支持计划
关键词
数字全息
三维位移场
统一设备架构
并行重建
digital holography
3D displacement field
compute unified device architecture
parallel reconstruction
