In computed tomography (CT) image reconstruction, the iterative algorithms have many advantages that the analytic algorithms don't have. However, the calculation of system matrix is time-consuming, which affects their application in real-time CT image reconstruction. In this paper, a fast system matrix calculation method is proposed on the basis of that the distance between the voxel and the adjacent rays reflects the contribution of the voxel to the ray projections. Firstly, every voxel is projected at a projection angle, and then the weighting factor is determined according to the distance between the voxel's projection and the adjacent ray projection on the detector. Finally, the ray equations and weighting factors corresponding to every voxel at all projection angles are obtained. The calculation amount of the proposed method are greatly improved compared with the current grid traversal (GT) method, and it has good parallel computing performance. In the simulation experiment, the Shepp-Logan head phantom is reconstructed by using the proposed system matrix calculation method and algebraic reconstruction technique (ART). Compared with GT method, its cost is reduced by nearly 30 %, and its acceleration ratio is four times that of the GT method. The experimental results show that it is a fast and effective method to calculate the system matrix. [ABSTRACT FROM AUTHOR]