High-attenuation objects like metals will result in metal artifacts in computed tomography images. Compared with single metallic object, artifacts due to multiple and large-scaled metallic objects is more complicated in representation and have much worse effects on reconstructed image. State-of-the-art metal artifacts reduction for multiple metal objects based on interpolation method cannot solve the beam hardening inside the metals, and can easily make mistakes in segmentation and interpolation. Aiming at reduction of multiple metallic objects, this paper simulates the production of the artifacts and proposes a metal artifacts reduction method based on projections correction. In this method, metal regions are firstly segmented directly from projection domain, and then a correction model is established for projections in metal regions. Finally, correction is made by adjusting parameters of the model. The optimal solution of the parameters is achieved by NM-simplex method that makes the gray entropy of the reconstructed image minimum. The simulation results and obtained data show that the present method significantly improves metal artifact due to multiple metallic objects and provides a better image quality than that obtained using interpolation.