Distinguishing real and virtual edge intersection in pairs of uncalibrated images

Abstract

Precise determination of object planes in images is very important in applications of computer vision, such as pattern recognition and 3D reconstruction. The corners of a polygonal object plane, e.g. roof, wall, etc., can be determined in an image by detecting and intersecting edge straight lines bounding the plane. Any two non-parallel lines in an image intersect at an image point. If this intersection corresponds to a 3D point in the scene, it is called a real intersection, otherwise it is called a virtual intersection. An automatic system for locating image lines is likely to produce many virtual intersections. This paper presents a computational technique to discriminate between real and virtual intersections. The method is based on rectified images obtained from a pair of uncalibrated images and is illustrated with images of a real scene. The results obtained showed reliable decisions. © Springer-Verlag Berlin Heidelberg 2000.