This work was done during the summer (2000) research stage, at INRIA, Sophia Antipolis, France under the supervision of Dr. Rachid Deriche (Research Director, Rachid.Deriche@sophia.inria.fr) and Dr. Theo Papadopoula (Researcher, Theodore.Papadopoula@sophia.inria.fr) in ROBOTVISION group headed by Olivier D. Faugeras.
3-D reconstruction of scenes from its 2-D photographic images is one of the fundamental research topics of Computer Vision. Self-Calibration is one of the first task encountered in this process when we do not know the intrinsic parameters of the Camera (focal length of lens and principal points) from which images have been taken. The latest works in this field include the work from Manolis and Rachid Deriche in which they use SVD technique to simplify and reduce the kruppa equations in the case when all images are of full perspective type (wide field-of-view images). Long Quan has worked on the Calibration of images from Affine cameras (narrow field-of-view/zoomed-in images/closeup). But no work had been done on the mixed case when both full perspective images (wide field-of-view images) and Affine cameras (narrow field- of-view/zoomed-in/closeup images) are present. In this paper we present a technique to deal with this new case.