Abstract: A beam hardening correction method based on indexing projection and weight function is proposed to remove the cupping artifacts of low-energy X-ray industrial CT images and to improve the identification capability and quantitative analysis. Firstly, it is concluded from theoretical analysis that cupping artifacts is dominated by beam hardening occurring when X-ray continuous spectrum passes through the object to be measured. Then, projection data and line attenuation coefficient of different thickness are acquired through scanning ladder model. The hardening model function and weight function correction model function which determines the power function are obtained by fitting curve. Furthermore, we scan the measured round workpiece to get the projection data under different indexing. Finally, for each index projection data, the projection data of cupping artifacts are corrected by combining the weight function and current indexing projection data. A correction experiment is carried out in the actual CT images including cupping artifacts. The result indicates that the SNR of gray image using the proposed method has increased by 3.29% with no amplification of noise, which shows that the proposed method more effectively eliminates the cupping artifacts and keeps better image edges as compared with the polynomial fitting method.