Institute Homepage EN Sign In

Back

Empirische Inferenz Conference Paper 2008

Stereo Matching for Calibrated Cameras without Correspondence

We study the stereo matching problem for reconstruction of the location of 3D-points on an unknown surface patch from two calibrated identical cameras without using any a priori information about the pointwise correspondences. We assume that camera parameters and the pose between the cameras are known. Our approach follows earlier work for coplanar cameras where a gradient flow algorithm was proposed to match associated Gramians. Here we extend this method by allowing arbitrary poses for the cameras. We introduce an intrinsic Riemannian Newton algorithm that achieves local quadratic convergence rates. A closed form solution is presented, too. The efficiency of both algorithms is demonstrated by numerical experiments.

Author(s): Helmke, U. and Hüper, K. and Vences, L.
Book Title: CDC 2008
Journal: Proceedings of the 47th IEEE Conference on Decision and Control (CDC 2008)
Pages: 2408-2413
Year: 2008
Month: December
Day: 0
Publisher: IEEE Service Center
Bibtex Type: Conference Paper (inproceedings)
Address: Piscataway, NJ, USA
DOI: 10.1109/CDC.2008.4738620
Event Name: 47th IEEE Conference on Decision and Control
Event Place: Cancun, Mexico
Digital: 0
Electronic Archiving: grant_archive
Institution: Institute of Electrical and Electronics Engineers
Language: en
Organization: Max-Planck-Gesellschaft
School: Biologische Kybernetik
Links:

BibTex 

@inproceedings{5633,
  title = {Stereo Matching for Calibrated Cameras without Correspondence},
  journal = {Proceedings of the 47th IEEE Conference on Decision and Control (CDC 2008)},
  booktitle = {CDC 2008},
  abstract = {We study the stereo matching problem for reconstruction of the location of 3D-points on an unknown surface patch from two calibrated identical cameras without using any a priori information about the pointwise correspondences. We assume that camera parameters and the pose between the cameras are known. Our approach follows earlier work for coplanar cameras where a gradient flow algorithm was proposed to match associated Gramians. Here we extend this method by allowing arbitrary poses for the cameras. We introduce an intrinsic Riemannian Newton algorithm that achieves local quadratic convergence rates. A closed form solution is presented, too. The efficiency of both algorithms is demonstrated by numerical experiments.},
  pages = {2408-2413},
  publisher = {IEEE Service Center},
  organization = {Max-Planck-Gesellschaft},
  institution = {Institute of Electrical and Electronics Engineers},
  school = {Biologische Kybernetik},
  address = {Piscataway, NJ, USA},
  month = dec,
  year = {2008},
  slug = {5633},
  author = {Helmke, U. and H{\"u}per, K. and Vences, L.},
  month_numeric = {12}
}