An epipolar rectification algorithm for uncalibrated images

Abstract

Traditional epipolar rectification algorithms require stereo images captured with binocular parallel stereo camera, and the requirements for the image acquisition system are high. In this paper an epipolar rectification algorithm for uncalibrated images which are not from binocular camera is proposed. Firstly, Minimum Enclosing Rectangle (MER) is calculated according to matching points, which are from Scale Invariant Feature Transform (SIFT); then image pairs adjustment is implemented with four vertex points parameters of MER; lastly, combining Hartley algorithm with Francesco algorithm epipolar rectification is realized. The experimental results show that the proposed algorithm is invariant to translation, rotation and scaling transformation to a certain extent. © 2012 Springer Science+Business Media, LLC.