Recovery of epipolar geometry from 2D point matches

Contents of this page

General description and related papers
Source code and documentation
Demos and performance graphs
History

General description and related papers

On this page you will find source code and demos to compute the epipolar geometry associated to a stereo pair of images. Given two images in which points of interests have been selected and put in correspondence, the proposed algorithm: This algorithm is specially interesting when a high proportion of outliers is expected (it has been shown to work for 90% outliers and more), or when the rigidity assumption (relevance of the epipolar geometry) is unsure (the algorithm tells how "rigid" the point correspondences are, by simultaneously taking into account the localization error and the number of matches; if no real stereo correspondence holds, it won't return one by accident).

For mode precise details about the method and the algorithm, see

  L. Moisan, B. Stival, "A probabilistic criterion to detect rigid point matches between two images and estimate the fundamental matrix", International Journal of Computer Vision, vol. 57:3, pp. 201-218, 2004.
  PostScript of preprint       PDF of preprint       bibtex entry


You may freely use the programs given below. If you do, please make reference to this web page and to the paper mentioned above. If you want to distribute the codes, please do not modify them (or if you do so, indicate clearly in the codes that you did so).

Source code and documentation

The source codes are written in C as MegaWave2 modules. Generic C versions should be available soon.

  stereomatch.c    documentation (PDF)    Detect and rate the best stereo correspondence between 2D point matches (main module)


Demos and performance graphs

work in progress...

History

october 2007: first version

back to L. Moisan home page