Calib3d

calib3d is a module of OpenCV that provides algorithms for three-dimensional computer vision. It focuses on camera calibration, stereo vision, and 3D reconstruction. The module supplies tools to estimate both intrinsic parameters of a camera and extrinsic relationships between cameras, enabling accurate measurement, scene understanding, and pose estimation from image data.

It supports calibrating cameras using calibration patterns such as chessboards or circle grids, computing distortion coefficients, and recovering focal lengths and principal points. For stereo setups, it offers stereoCalibrate to estimate both cameras’ parameters and the relative pose, followed by stereoRectify to produce rectified image pairs. It also provides methods for undistortion and rectification, depth from disparity through disparity algorithms like StereoBM and StereoSGBM, and reprojectImageTo3D to obtain 3D coordinates from disparity maps. In epipolar geometry, calib3d provides functions to compute the fundamental and essential matrices and to recover camera motion with recoverPose or solvePnP variants for pose estimation from 2D-3D data.

Key utility functions include calibrateCamera, stereoCalibrate, stereoRectify, initUndistortRectifyMap, remap, undistortPoints, getOptimalNewCameraMatrix, and Rodrigues. Projection and reconstruction tasks are supported through projectPoints and triangulatePoints, while pose estimation uses solvePnP and solvePnPRansac. The module is available in C++ and Python APIs and is widely used in robotics, augmented reality, and photogrammetry to calibrate cameras, rectify imagery, and reconstruct 3D structure from image data.