Share Email Print
cover

Proceedings Paper

Wide baseline stereo matching based on double topological relationship consistency
Author(s): Xiaohong Zou; Bin Liu; Xiaoxue Song; Yang Liu
Format Member Price Non-Member Price
PDF $14.40 $18.00

Paper Abstract

Stereo matching is one of the most important branches in computer vision. In this paper, an algorithm is proposed for wide-baseline stereo vision matching. Here, a novel scheme is presented called double topological relationship consistency (DCTR). The combination of double topological configuration includes the consistency of first topological relationship (CFTR) and the consistency of second topological relationship (CSTR). It not only sets up a more advanced model on matching, but discards mismatches by iteratively computing the fitness of the feature matches and overcomes many problems of traditional methods depending on the powerful invariance to changes in the scale, rotation or illumination across large view changes and even occlusions. Experimental examples are shown where the two cameras have been located in very different orientations. Also, epipolar geometry can be recovered using RANSAC by far the most widely method adopted possibly. By the method, we can obtain correspondences with high precision on wide baseline matching problems. Finally, the effectiveness and reliability of this method are demonstrated in wide-baseline experiments on the image pairs.

Paper Details

Date Published: 6 August 2009
PDF: 9 pages
Proc. SPIE 7384, International Symposium on Photoelectronic Detection and Imaging 2009: Advances in Imaging Detectors and Applications, 738405 (6 August 2009); doi: 10.1117/12.832371
Show Author Affiliations
Xiaohong Zou, Yanshan Univ. (China)
Bin Liu, Yanshan Univ. (China)
Xiaoxue Song, Yanshan Univ. (China)
Yang Liu, Yanshan Univ. (China)


Published in SPIE Proceedings Vol. 7384:
International Symposium on Photoelectronic Detection and Imaging 2009: Advances in Imaging Detectors and Applications
Kun Zhang; Xiang-jun Wang; Guang-jun Zhang; Ke-cong Ai, Editor(s)

© SPIE. Terms of Use
Back to Top