Camera Calibration

Camera calibration involves estimating camera parameters (including intrinsic and extrinsic parameters) to infer geometric features from captured image sequences, which is a critical step in computer vision and robotics. Based on different neural network architectures, researchers have developed two primary learning-based camera calibration paradigms: regression-based calibration and reconstruction-based calibration. These methods achieve high-precision camera calibration by directly predicting parameter values and optimizing the 3D scene reconstruction, respectively, thereby enhancing the performance and robustness of visual systems.