针对闪光照相图像信噪比低的特点，提出了一种改进的约束共轭梯度闪光照相图像重建算法。该算法在约束共轭梯度迭代重建的基础上，提出了新的预优矩阵选取方案，减小了重建图像的轴线噪声，利用松弛迭代步长代替共轭梯度法中的最优迭代步长，并采用了新的收敛准则，在保证算法收敛的同时，减少了重建算法的计算量。数值试验表明，与传统约束共轭梯度重建算法相比，改进算法稳定收敛，迭代速度更快，并能有效提高重建质量。

This paper presents an improved constrained conjugate gradient reconstruction algorithm for flash radiographic image with low signal-to-noise ratio. A new preconditioned method is consummated which can more accurately reconstruct image’s axis of symmetry. The new convergence criterion and the substitution of relaxation step for optimal step in the one-dimension search are adopted to ensure the reconstruction accuracy and shorten the reconstruction time simultaneously. The numerical experiment results show that the improved algorithm has better general performance in iteration time and reconstruction quality compared with the original constrained conjugate gradient algorithm.