代数重建中一种投影系数的快速计算方法

代数重建过程中,投影系数的计算是影响重建速度的主要因素。本文从射线与网格相交的规律出发,先用一个数组存储投影系数值,再使用一个数组修正,通过两个数组相结合来求解射线穿过网格的投影系数。该算法不仅减少了求解投影系数的计算量,并且抑制了在程序设计时所需的分支结构,使其适合于CUDA并行加速。该算法在重建固体火箭发动机模型中,在保证图像质量不损失的前提下,在CPU端与Siddon算法相比,重建速度提高了10倍多。

Fast algorithm for calculating the radiological path in fan-beam CT image reconstruction

The calculation of projection coefficients is the key point to iterative reconstruction algorithms, and it mainly affects the reconstructed quality and speed. This paper starts from the rule of rays intersect with the grid, uses an array to store the projection coefficient value firstly, then uses an array to correct it, and solves the ray projection coefficient through the grid by combining the two arrays. The proposed method could clearly reduce the program calculation, effectively remove redundant switch branches, which simplifies the whole process so that it"s suitable for using compute unified device architecture (CUDA) to parallel speedup. On the premise of guarantee the quality of the image is not loss, the time usage of reconstructing a solid rocket engine simulator by use of the proposed algorithm is compared with that of Siddon algorithm in central processing unit (CPU). Experimental results show that the proposed algorithm is 10 times faster than Siddon algorithm.