Robust camera calibration with vanishing points

Using vanishing points to calibrate camera is versatile and attracting. But it is still hard to be used in computer vision metrology or other fields striving for high accuracy. In this paper, we present a robust scheme for camera internal parameters estimation using pairs of orthogonal vanishing points. RANSAC scheme is used to eliminate the outliers caused by degeneracy or gross noise, and in each iteration an isotropic normalization is applied to the input vanishing points coordinates to make the estimation more accurate and stable. Experiment results with synthetic camera images validate the accuracy and robustness of the proposed algorithm.