Paracatadioptric camera calibration using sphere images and common self-polar triangles

Abstract

Camera calibration is an indispensable step in achieving stratified three-dimensional reconstruction. In the imaging model of the unit viewing sphere of a central catadioptric camera, there exists a pair of common self-polar triangles between a pair of antipodal sphere images and the image of the great circle on the base plane, so the vertices of the common self-polar triangles can be obtained. According to the imaging characteristics of the sphere on the unit viewing sphere, the line connecting the two common poles outside the conics is a vanishing line, enabling the imaged circular points to be determined. Finally, the intrinsic parameters of the paracatadioptric camera can be found using constraints between the imaged circular points and the image of the absolute conic. Since the boundary points of the sphere image can be almost completely extracted, the curve fitting accuracy can be improved. Experiments with simulated as well as real data demonstrated that our method is feasible and effective.