【作者】 江鹏； 陈志强； 邢宇翔；

【Author】 JIANG Peng CHEN Zhi-qiang XING Yu-iang (Department of Engineering Physics,Tsinghua University,Beijing 100084,China)

【机构】 清华大学工程物理系；

【摘要】 三维锥束CT图像的FDK算法重建由于运算量大,在重建高分辨率的图像时,重建所需时间通常难以满足实际应用的需求。集群并行计算是解决上述问题的常用方法之一,通常采用的方法是基于不同角度投影数据的无关性实现按角度划分投影数据实现并行。而另一种思路通过对重建对象的分层无关性也可以实现重建的并行处理。本文针对三维锥束CT 中常用的FDK 算法,研究了按以上两种思路实现的不同并行化方法,并给出了在一个小的集群系统上的模拟实验结果。更多还原

【Abstract】 Image reconstruction of 3D cone-beam tomography using FDK algorithm requires hugecomputational resource.The reconstruction of a highresolution 3D object would usually take a long time,which is intolerable in real application.Parallel computing using cluster systems is one of the commonapproaches to solve this problem.Generally,the parallelization is implemented by separating theprojections according to angles due to the irrelevance between the projection data from each angle.Another method is to divide an object into slices which can be reconstructed simultaneously.Thispaper described these two parallelization methods of the FDK algorithm.The results of our simulationson a small cluster system is reported.更多还原