Euclidean and Affine Structure/Motion for Uncalibrated Cameras from Affine Shape and Subsidiary Information

The paper deals with the structureemotion problem for un-calibrated cameras, in the case that subsidiary information is available, consisting e.g. in known coplanarities or parallelities among points in the scene, or known positions of some focal points (hand-eye calibration). Despite unknown camera calibrations, it is shown that in many instances the subsidiary information makes aane or even Euclidean reconstruction possible. A parametrization by aane shape and depth is used, providing a simple framework for the incorporation of apriori knowledge , and enabling the development of iterative, rapidly converging algorithms. Any number of points in any number of images are used in a uniform way, with equal priority, and independently of coordinate representations. Moreover, occlusions are allowed.