Abstract: Three-dimensional imaging analysis for orthodontics remains underdeveloped, while two-dimensional imaging suffers from blurring and distortion due to dimensional limitations. This study proposed a cone-beam computed tomography (CBCT) reorientation based on 3D landmarks and 2D cephalometry simulation visualization method. First, a deep learning approach predicted specific landmarks on 3D images. A reference coordinate system was established according to the correspondence between these landmarks and head posture, and CBCT data was reoriented to achieve standardized head alignment. Next, an algorithm simulated the imaging process of a 2D cephalometric radiograph to produce rendered images. Then, through post-processing methods such as including nonlinear mapping and regional enhancement, generated simulated 2D cephalometric images suitable for observation and measurement. The proposed method achieved a reorientation Euler angle error of mean 0.8° and standard deviation 0.5° on a non-public dataset of 84 CBCT scans. The generated simulated images can be directly used in existing manual and automatic 2D cephalometric analysis methods. In a user study evaluating seven quality metrics, 80% of respondents rated the simulated images superior to conventional 2D radiographs.