Summary: A Three­view Matching Algorithm Considering Foreshortening Effects
Ning Xu and Narendra Ahuja
Beckman Institute, University of Illinois at Urbana­Champaign
Abstract
In this paper, we present a three­view matching algorithm
that takes into account surface foreshortening effects to
match image feature points and provides surface normal
directions at these points as a by­product. Compared to
other existing, similar algorithms that consider surface
foreshortening effects, our algorithm is more efficient
since it decomposes a 2D surface orientation search into
two 1D searches along two pairs of epipolar lines, and
more robust since two pairs of stereo images from three
cameras are used and the decomposition makes it possi­
ble for our algorithm to detect and match surface points
near 3D edges. Experimental results showing correspon­
dences, surface normal directions, reconstructed surfaces
are presented and evaluated.
Introduction
The task of matching points between two images is cen­

Source: Ahuja, Narendra - Department of Electrical and Computer Engineering, University of Illinois at Urbana-Champaign

Collections: Computer Technologies and Information Sciences; Engineering