中文摘要：

外部计算机断层成像（CT）重建主要用于重建管状物（如管道）的管壁截面图像,具有重要的实际意义。外部CT重建算法中,带子区域平均图像总变差最小的凸集投影（SA-TVM-POCS）重建数值算法可以得到高质量的CT图像,可用于管道的无损检测。但由于在实现过程中计算量较大,阻碍了其在一些时间要求较高环境中的应用。计算机统一设备架构（Computer Unified Device Architecture,CUDA）是近几年发展起来的解决高强度计算的有效工具之一,该文利用CUDA的高强度并行计算性能,提高外部CT问题SA-TVM-POCS算法的速度,从而增加该算法的应用范围;为充分利用CUDA其高强度数据级并行计算能力,该文改进了SA-TVM-POCS算法的实现过程以适应CUDA的并行计算架构。实验结果表明,该文方法可以在重建图像质量没有下降的情况下,加速比达到20倍以上;所以,利用CUDA可以加快SA-TVM-POCS算法的计算速度。

英文摘要：

The exterior Computed Tomography（CT） reconstruction majors in reconstructing the cross section image of pipe wall in reality.Subset Average-Total Variation Minimization-Projection Onto Convex Sets（SA-TVM-POCS） numerical reconstruction algorithm is one of the algorithms of exterior reconstruction methods that is stabilization and has good quality reconstructed image on non-destructive-testing of pipes.One disadvantage of the SA-TVM-POCS algorithm,just as other iterative algorithms,is the long computation time typically associated with reconstruction.This obstacle can be overcomed by implementing reconstruction algorithms on Computer Unified Device Architecture（CUDA）.In this work the SA-TVM-POCS algorithm is implemented by CUDA in order to signi？cantly reduce the reconstruction time.The mathematical and computational details of the implementation are explored.The experiment results show that the algorithm can be speedup to 20 times without reducing the quality of reconstruction image.So,the SA-TVM-POCS algorithm of exterior computed tomography can be accelerated by CUDA effectively.