Matching feature points for telerobotics

A system that quickly and reliably matches points from images of the same scene, taken by three different cameras, is presented. The first step consists of weak calibration of the camera system. Feature points are then iteratively selected and matched in the first two images, guided by a fundamental matrix. Each correspondence is then validated by computing the position of matches in the third image using a trifocal tensor, and enforcing a constraint on the disparity of neighboring matches.