Robust Fundamental Matrix Determination without Correspondences

Lehmann, Stefan, Clarkson, I. Vaughan L., Bradley, Andrew P., Williams, John and Kootsookos, Peter J. (2005). Robust Fundamental Matrix Determination without Correspondences. In: Brian C. Lovell and Anthony J. Maeder, Proceedings of the APRS Workshop on Digital Image Computing, Pattern Recognition and Imaging for Medical Applications. APRS Workshop on Digital Image Computing (WDIC2005), Griffith University, Southbank, Brisbane Australia, (97-102). 21 February, 2005.

Attached Files (Some files may be inaccessible until you login with your UQ eSpace credentials)
Name Description MIMEType Size Downloads
n14.pdf n14.pdf application/pdf 176.29KB 481

Author Lehmann, Stefan
Clarkson, I. Vaughan L.
Bradley, Andrew P.
Williams, John
Kootsookos, Peter J.
Title of paper Robust Fundamental Matrix Determination without Correspondences
Conference name APRS Workshop on Digital Image Computing (WDIC2005)
Conference location Griffith University, Southbank, Brisbane Australia
Conference dates 21 February, 2005
Proceedings title Proceedings of the APRS Workshop on Digital Image Computing, Pattern Recognition and Imaging for Medical Applications
Place of Publication St Lucia, Qld.
Publisher The University of Queensland
Publication Year 2005
Sub-type Fully published paper
ISBN 0-9580255-3-3
Editor Brian C. Lovell
Anthony J. Maeder
Volume 1
Issue 1
Start page 97
End page 102
Total pages 6
Collection year 2005
Abstract/Summary Estimation of the fundamental matrix is key to many problems in computer vision as it allows recovery of the epipolar geometry between camera images of the same scene. The estimation from feature correspondences has been widely addressed in the literature, particularly in the presence of outliers. In this paper, we propose a new robust method to estimate the fundamental matrix from two sets of features without any correspondence information. The method operates in the frequency domain and the underlying estimation process considers all features simultaneously, thus yielding a high robustness with respect to noise and outliers. In addition, we show that the method is well-suited to widely separate viewpoints.
Subjects 280208 Computer Vision
280203 Image Processing
280204 Signal Processing
E1
Keyword iris-research
structure and motion
fundamental matrix
epipolar geometry
projection-slice theorem
orthographic projection
feature correspondences
Q-Index Code E1

 
Versions
Version Filter Type
Citation counts: Google Scholar Search Google Scholar
Created: Mon, 28 Feb 2005, 10:00:00 EST by Stefan Lehmann on behalf of School of Information Technol and Elec Engineering