康普顿背散射图像重建算法的GPU加速研究  

GPU Acceleration of Compton Backscatter Image Reconstruction Algorithm

作　　者：李守鹏[1] 闫镔[1] 刘拥军[1] 李磊[1] 

机构地区：[1]信息工程大学信息系统工程学院,河南郑州450002

出　　处：《信息工程大学学报》2012年第6期713-717,共5页Journal of Information Engineering University

基　　金：国家863计划资助项目(2012AA011603)

摘　　要：在康普顿背散射成像中,射线源和探测器放置于物体同侧,无需旋转即可重建出物体内部断层图像。但康普顿背散射图像重建算法计算复杂度较高,重建耗时较长。根据康普顿背散射图像有序子集最大期望法(ordered subsets expectation maxization,OSEM)重建算法的执行特点,设计了适合算法正投影和反投影的并行策略进行GPU加速。重建结果表明,所使用的加速策略能够在保证重建图像质量的同时获得较高的加速比。In Compton backscatter imaging, the ray source and detectors are placed at the same side of object; we can reconstruct the object without rotation process. But Compton backscatter image re- construction algorithm suffers high computation cost and long reconstruction time. According to the characteristics of OSEM reconstruction algorithm, we design proper parallel strategies for projection process and backprojection process to accelerate the algorithm. The results show that using these parallel strategies we can obtain a high accelerating ratio as well as a high quality image.

关 键 词：康普顿背散射 图像重建 OSEM GPU加速 CUDA 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]