Tutorial on Homographies Markelic（教程单应性Markelic）.pdfVIP

Tutorial on Homographies Irene Markeli´c Bernstein Center for Computational Neuroscience, University of G¨ottingen, Bunsenstrasse. 10, 37073 G¨ottingen, Germany {irene,woergoetter}@bccn-goettingen.de http://www.bccn-goettingen.de Abstract. Key words: homographies 1 Introduction This text aims at explaining homographies and to give a practical guide on how to use it. A homography is a transformation from one Projective plane to another. This is very useful for generating ”fake” images on objects from views that were never actually used. Also inverse perspective mapping, IPM, makes use of this, by projecting a taken image on a know plane. This way one gets rid of the perspective effect on objects that are on that plane, for all others this transformation results in a distorted view. To understand homographies some background knowledge about the Projective space and camera modeling is necessary. This tutorial is not complete, yet. So far, it covers most of the necessary background knowledge and is to be extended. 2 Projective Geometry Projective geometry is an extension of the familiar Euclidean geometry. The latter allows us to describe transformations like translations and rotations, but we cannot model more complicated transformations like perspective (or cen- tral) projections. Perspective projection is a mathematical model (description) that allows us to precisely describe how the world around us gets mapped on a plane, it is described in the section ”Perspective Projection” below. Note, that there are also other projections, e.g. the orthographic projection, but we will not consider those here. Why can we not use Euclidean geometry for this transforma- tion of perspective proje