A new evaluation criterion for point correspondences in stereo images

In this paper, we present a new criterion to evaluate point correspondences within a stereo setup. Many applications such as stereo matching, triangulation, lens distortion correction, and camera calibration require an evaluation criterion, indicating how well point correspondences fit to the epipolar geometry. The common criterion here is the epipolar distance. Since the epipolar geometry is often derived from noisy and partially corrupted data, an uncertainty regarding the estimation of the epipolar distance arises. However, the uncertainty of the epipolar geometry, in the shape of the covariance matrix of an epipolar line, provides additional information, and our approach utilizes this information for a new distance measure. The basic idea behind our criterion is to determine the most probable epipolar geometry that explains the point correspondence in the two views. Furthermore, we show that using Lagrange multipliers, this constrained minimization problem can be reduced to solving a set of three linear equations.