针对常规 CT系统中 X射线是连续宽能谱而导致重建图像中衰减特性相近组分对比度较低的问题，论文提出一种分解多电压 X射线图像序列获取新投影的方法，用新投影重建的图像有效地提高了对衰减特性相近组分的区分能力。该方法依据 X射线成像规律，将各电压下图像看作多个窄能谱图像的叠加，借鉴非负矩阵分解，以分解的误差平方和最小为目标对多电压 X射线图像分解。用硅铝材质的圆柱体实验，新投影重建的图像相比直接重建图像，硅与铝对比度明显提高，表明了方法的有效性。

The X-ray tube of usual computed tomography imaging system produces X-ray beam with a polychromatic spectrum, which results in low contrast of the materials with approximate attenuation coefficients in the reconstructed image. A method was proposed to obtain a new projection by decomposition of multi-voltage X-ray images. The distinction of the materials was enlarged in the reconstructed image by the new projection. In this method, the X-ray image of each voltage was considered as a sum of many narrow-energy-width X-ray images according to the law of X-ray imaging. Referring to non-negative matrix factorization, the new projection was got by minimizing the sum of squared error of multi-voltage X-ray images’ decomposition. A cylinder composed of silicon and aluminum was used in the experiment. The contrast of silicon and aluminum is evidently improved in the new reconstructed image compared to the direct reconstructed image. It shows the effectiveness of the proposed method.