Robust camera calibration with epipolar constraints

In this paper, we present a robust approach based on epipolar constraints for performing camera calibration. Contrary to the poor performance of existing calibrating methods in noise situations, the proposed technique defines a deletion criterion using epipolar constraints and maximum likelihood estimation and deletes the control points which are significantly disturbed by image noise or badly located by the corner detector, before the camera parameters are finally computed. Experiments show that our approach demonstrates good performance in terms of accuracy, convergence and robustness even with various perturbed reference points.