LI Bei-bei, LI Zhi-hua, BAN Xiao-zheng. Parallelization method for filtering accelerating of convoluted back projection algorithm based on GPU[J]. 2019, (4): 69-72.
LI Bei-bei, LI Zhi-hua, BAN Xiao-zheng. Parallelization method for filtering accelerating of convoluted back projection algorithm based on GPU[J]. 2019, (4): 69-72. DOI: 10.13873/J.1000-9787(2019)04-0069-04.
GPU加速卷积反投影算法的滤波并行化方法
摘要
当重建的图像规模偏大、实时性要求高时
卷积反投影(CBP)重建过程比较慢
达不到预期满意的速度。针对这一不足
通过深入研究卷积反投影算法的原理
优化投影数据在图形处理器(GPU)中的存储结构、分析和挖掘算法执行过程中滤波阶段的可并行性
对其中的滤波操作进行并行化处理
从而提出并行滤波过程的方法。通过在MATLAB进行仿真实验
实验结果表明:所提出的并行化方法在保证重建图像精度和清晰度的前提下
同串行卷积法相比较
滤波过程运算的加速比得到了较大程度的提高。
Abstract
Convoluted back-projection( CBP) algorithms is a kind of image reconstruction algorithm
it have been applied widely in medical image and non-destructive testing
but its reconstruction speed is far from satisfying in dealing large image. According to the principle of traditional CBP algorithms and parallelism in the filtering process of graphic processing unit( GPU)
adopts the parallel filtering process of GPU
meanwhile
optimize internal data structure. In evaluating the the capability of the process
applies the contrast experiment between direct convoluted method. The following result show that speed of convoluted back-projection filtering process is accelerated to a large extent compared to direct convoluted method.