distCoeffs

distCoeffs is a container for the distortion coefficients of a camera lens used in OpenCV’s camera calibration and undistortion workflows. These coefficients describe how an ideal pinhole projection is distorted by the lens and sensor, and they are used by functions such as undistort, undistortPoints, and initUndistortRectifyMap to map distorted image points back to their undistorted locations.

In OpenCV, distCoeffs is stored as a 1xN or Nx1 vector (or a 1D array in higher level bindings) of floating point values. The length N depends on the distortion model selected during calibration and commonly takes the values 4, 5, 8, or 12. The coefficients typically encode radial distortion, tangential distortion, and optional higher-order terms. A common ordering is as follows: k1, k2, p1, p2, k3, k4, k5, k6, s1, s2, s3, s4, where k1, k2, k3 (and optionally k4–k6) model radial distortion, p1 and p2 model tangential distortion, and s1–s4 model thin prism distortion when present. The exact interpretation depends on the calibration model chosen.

Calibration routines produce distCoeffs as part of a camera model for a given lens. They depend on the intrinsic camera matrix and the calibration data. Practically, distCoeffs is used together with the camera matrix to remove distortion from images and to compute rectified image mappings. In Python, it appears as a 1xN numpy array; in C++ it is a cv::Mat.