Robust subpixel stereo matching by relaxation of match candidates

A key issue in performing quantitative measurements in stereoscopic images is the problem of establishing stereo correspondence. For this purpose a novel two step algorithm is presented. The commonly neglected problem of ambiguities is addressed and a solution based on, the relaxation of match candidates is introduced. The method comes with the major benefits of a reverse matching step at almost no additional computational cost and requires no initial estimations or constraints. The results of both match directions are re-used for a robust subpixel refinement. Experiments are performed on ground truth and a comparison to other well known techniques is presented. The method performs best on real images, but not tinder all of the used noise conditions.