Manufacturing Multiple View Constraints
David McKinnon, Brian C. Lovell · 2005
In this paper we present an algorithm for the gen-eration of the multiple view constraints for arbitrary configurations of cameras and image features corre-spondences. Multiple view constraints are an impor-tant commodity in computer vision since they facilitate in determining camera locations using only the corre-spondences between common features observed in sets of uncalibrated images. We show that by a series of counting arguments and a systematic application of the principles of antisymmetric algebra it is possible to gen-erate arbitrary multiple view constraints in a completely automated fashion. The algorithm has already been uti-lized to discover new sets of multiple view constraints for surfaces. 1.