Robust Relative Pose Estimation of Two Cameras by Decomposing Epipolar Geometry (Paperback)

This work approaches the image-based estimation of two cameras' relative pose by a novel decomposition of epipolar geometry. It gives consistent insight into the composition of the essential and the fundamental matrix. Furthermore, the decomposition enables the relative pose problem to be split into an outer and inner optimization problem in which the two components of epipolar geometry, the epipoles and the epipolar line homography, can be estimated separately with the minimal number of parameters. Two possible approaches for relative pose estimation have been developed in which the proposed factorization could be applied.