Abstract:
When the resolution of reconstruction image is high, the FDK algorithm of cone-beam CT image reconstruction is time-consuming. To improve the time performance of the FDK algorithm, two parallelization algorithms are analyzed. Claster parallel computing and SSE instruction optimization technology are applied to the FDK algorithm and the algorithm is realized on an 8 nodes cluster. Results show that the reconstruction speed of the image with 2563 resolution is improved about 29 times via this method.
Key words:
Computed Tomography(CT),
FDK algorithm,
parallel computing,
cluster,
instruction optimization
摘要: 为提高锥束CT的FDK重建算法在重建高分辨率的图像时的速度,分析2种并行策略及其对应的通信时耗,研究集群并行与SSE指令优化计算相结合的FDK算法,在8个节点的集群系统上进行实现。实验结果表明,采用集群并行加指令优化的方式,可将分辨率为2563的图像的重建速度提高到原来的29倍。
关键词:
计算机断层成像,
FDK算法,
并行计算,
集群,
指令优化
CLC Number:
ZOU Yong-ning; LIU Bao-dong;. FDK Reconstruction Algorithm Based on Cluster Parallelization and Instruction Optimization[J]. Computer Engineering, 2009, 35(8): 10-12.
邹永宁;刘宝东;. 基于集群并行及指令优化的FDK重建算法[J]. 计算机工程, 2009, 35(8): 10-12.