1. Introduction
Structure from Motion (SfM) employs a set of 2D images acquired by a moving camera to estimate the 3D geometry of a scene and the camera motion. The vast majority of relevant research assumes that images have been acquired with ordinary pinhole cameras, which collect converging light rays. Hence, most existing SfM frameworks cannot be directly applied to generalised cameras, i.e. cameras which do not share a single centre of projection [9], [37].