摘要
在数字全息的再现过程中,需要对大量相位包裹图像进行解包裹处理,因此加快解包裹算法的运行速度是实现数字全息实时成像应用的关键。研究了基于GPU的快速相位解包裹算法。将无权最小二乘解包裹算法的主要计算过程转移到GPU中并行运算,并记录解包裹程序所需的时间。研究结果表明,与CPU独立运算相比,基于GPU的最小二乘解包裹算法运行速度得到了显著地提高,对于尺寸为512pixels×512pixels的包裹相位图,解包裹速度提高了10倍。
In order to realize the real-time analysis and observation of dynamic objects, the speed of digital reproduction must be considered in digital holography. A large number of wrapped phase images must be unwrapped in the reconstruction process. To accelerate phase unwrapping is the key to digital holographic real-time imaging application, which can greatly shorten the time of reconstructing phase images. An algorithm of speeding up phase unwrapping algorithm is studied by using heterogeneous computing with CPUq-GPU(Graphic Processing Unit). The main computational process of the un- weighted least square unwrapping algorithm is transferred to GPU. And the algorithm records the running time. It is found that GPU improves the computational efficiency of unwrapping significantly comparing with CPU. The speed of unwrapping the image with 512 pixels× 512 pixels increases ten times.
出处
《光学与光电技术》
2015年第4期23-27,共5页
Optics & Optoelectronic Technology
基金
广东省科技计划(2012B040302002)资助项目
关键词
全息
相位解包裹算法
图形处理器
并行计算
holography
phase unwrapping algorithm
graphic processing unit
parallel computation