This paper aims to solve the problem of preoperative computed tomography (CT) and intraoperative digitally subtracted angiograms (DSA) registration and proposes an automatic registration algorithm that synthesizes the feature information and grayscale information of medical images, which is characterized by high efficiency and accuracy. Firstly, this paper combines digitally reconstructed radiograph (DRR) technology and the Maximum Intensity Projection (MIP) principle to improve the efficiency of DRR image generation. Then, the DSA image feature terms are extracted using the Distance Transform Feature (DTF) and registered with the DRR image. The registration matrix is obtained by iteratively finding the extreme value of the similarity measure through Powell’s algorithm. The performance of the CT and DSA registration validation system is tested on the body membrane in real scenes. The results show that the 3D-2D image registration system based on MIP-DRR and DSA feature terms has a success rate of 85% in the head modality experiments, the registration fusion accuracy is not more than 0.7 mm, and the registration time is about 200 s. Conclusions: The 3D-2D medical image registration algorithm based on MIP-DRR and DSA feature terms has high accuracy, robustness, and registration efficiency.