›› 2015, Vol. 28 ›› Issue (10): 4-.

• 论文 • 上一篇    下一篇

图像代数重建算法的并行性研究

粟子芩,徐伯庆,苏栋骐   

  1. (上海理工大学 光电信息与计算机工程学院,上海 200093)
  • 出版日期:2015-10-15 发布日期:2015-10-29
  • 作者简介:粟子芩(1991—),女,硕士研究生。研究方向:图像重建。E-mail:suziqin2013@163.com。徐伯庆(1958—),男,博士,副教授。研究方向:通信及图像处理。

Research on Parallelism of Image Algebraic Reconstruction Algorithm

SU Ziqin,XU Boqing,SU Dongqi   

  1. (School of Optical-Electrical and Computer Engineering,University of Shanghai for Science and Technology,Shanghai 200093,China)
  • Online:2015-10-15 Published:2015-10-29

摘要: 针对CT图像重建中不完备投影数据重建问题,提出了基于迭代的代数重建(ART)算法。为了解决ART算法重建速度慢这一问题,文中采用计算机集群的并行加速处理技术,将投影角度均匀地分配给适当的多个CPU,再将多个CPU并行独立运行的重建结果进行加权平均。实验数据表明,图像重建速度提高近6倍(CPU个数为6),图像重建质量优于传统的串行重建质量,且提高了ART算法的实现效率。

关键词: 并行化加速;ART算法;图像重建;计算机集群

Abstract: The algebraic reconstruction technique (ART) is proposed to reconstruct image from incomplete projection data.Using the PC cluster system,we accelerate the ART reconstruction process based on dispatching projection angle evenly to each CPU and then average all the results.Experiments show that the speedup factor is roughly equal to the number of CPU's used appropriately and that the image quality is better than those by the classical serial algorithm.

Key words: parallel acceleration;algebraic reconstruction algorithm;image reconstruction;PC cluster

中图分类号: 

  • TP391.41