Determination of Ego-Motion from Matched Points

We propose an algorithm for the estimation of the motion of a camera moving through a static environment (ie. the ego-motion) from matched points on two images. The algorithm correctly weights the observations by minimising point mis-match distances on the image-plane. Prior knowledge concerning the camera motion may also be included. The algorithm is iterative, but generally converges quickly. Results are both more accurate and more robust than closed-form solutions based on the 8-Point Algorithm [Longuet-Higgins 1981, Faugeras 1987].