Camera Calibration Using Line Correspondences

In this paper, a method of determining the essential matrix for uncalibrated cameras is given, based on line matches in three images. The three cameras may have different unknown calibrations, and the essential matrices corresponding to each of the three pairs of cameras may be determined. Determination of the essential matrix for uncalibrated cameras is important, forming the basis for many algorithms such as computation of invariants image rectification, camera calibration and scene reconstruction. In the case where all the three cameras are assumed to have the same unknown calibration, the method of Faugeras and Maybank ([3, 4]) may be used to calibrate the camera. The scene may then be reconstructed exactly (up to a scaled Euclidean transformation). This extends previous results of Weng, Huang and Ahuja ([12]) who gave a method for scene reconstruction from 13 line correspondences using a calibrated camera. The present paper shows that the the camera may be calibrated at the same time that the scene geometry is determined. The research described in this paper has been supported by DARPA Contract #MDA97291-C-0053