利用CUDA技术实现锥束CT图像快速重建  

Fast Cone-beam CT Reconstruction with CUDA

作　　者：王珏[1] 曹思远[2] 邹永宁[1] 

机构地区：[1]重庆大学光电技术及系统教育部重点实验室ICT研究中心,重庆400044 [2]重庆大学自动化学院,重庆400044

出　　处：《核电子学与探测技术》2010年第3期315-320,共6页Nuclear Electronics & Detection Technology

基　　金：国家"863"计划(2006AA04Z104)资助项目

摘　　要：锥束CT三维重建算法的计算量和传输量巨大,仅利用CPU来计算,无法满足实时、快速、准确重建的要求,根据图形处理器运算能力强、存储带宽大的特点,研究了一种不需要学习图形API,就可以在图形处理器上实现三维重建算法的快速运算的方法。该方法采用基于统一计算设备架构的图形处理器,通过这种新架构的编程模式,利用图形处理器中的流处理器来加快滤波和反投影计算,实现了FDK算法的重建加速,与利用图形API的重建方法相比,开发门槛较低。对于尺寸为5123的单精度浮点数据格式的图像,重建时间可以缩短到一分钟以内,并且GPU与计算机的传输时间小于1秒。实验结果表明与仅利用CPU的重建方法相比,本文提出的图像加速方法得到了较高的时间加速比。Due to large in computation and transmission of cone-beam CT 3D reconstruction algorithm, it is impossible to meet the requirements of 3D image reconstruction in real-time, rapid and accurate by the means of using CPU only. The paper advances a method, without learning graphics API, to achieve the fast computing of algorithm by using graphics processor which has strong operation capability and large memory bandwidth. The method uses the kind of GPU based on CUDA, through new programming model, accelerating the filtering and backprojection by the Stream Processor Unit （SPU） in GPU, to achieve the FDK algorithm speed-up. Compared with the means using CPU only, the method is simpler in devel- opment than before. Experiment show that the image of 512^3 volume can be completed with 32bit float- ing-point in less than one minute, and the transmission time between the GPU and computer is less than one second. The experiment shows that the method gets a faster performance and good quality comparing with the method using CPU.

关 键 词：锥束CT 图形处理器 统一计算设备架构 流处理器FDK算法 重建加速 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]