Stereo autocalibration from one plane

Abstract

This paper describes a method for autocalibrating a stereo rig. A planar object performing general and unknown motions is observed by the stereo rig and, based on point correspondences only, the autocalibration of the stereo rig is computed. A stratified approach is used and the autocalibration is computed by estimating first the epipolar geometry of the rig, then the plane at infinity Π∞ (affine calibration) and finally the absolute conic Ω∞ (Euclidean calibration). We show that the affine and Euclidean calibrations involve quadratic constraints and we describe an algorithm to solve them based on a conic intersection technique. Experiments with both synthetic and real data are used to evaluate the performance of the method.