Accurate camera calibration using iteratively optimized algorithm of control points

Camera calibration and image correction is an important step for most of the computer vision tasks, like 3D metric measurement and multi-view stereo. In this paper, we propose a novel algorithm for camera calibration and image correction. It considers the inaccuracy of 3D pattern (like chessboard) and the lens distortion effect on the feature point detection. After each iteration, the estimated parameters are used to undistorted the images to have better feature detection. The established 3D-2D points are more accurate for further camera calibration. A comparative study based on synthetic data, real image and data from our 3D scanner corrupted by reasonable levels of noise shows that the proposed algorithm successfully calibrated, corrected the distorted images and precisely reconstructed the captured laser stripes in 3D space.