Richard Szeliski and Phil Torr
Structure from motion algorithms typically do not use exter- nal geometric constraints, e.g., the coplanarity of certain points or known orientations associated with such planes, until a final post-processing stage. In this paper, we show how such geometric constraints can be incorporated early on in the reconstruction process, thereby improving the quality of the estimates. The approaches we study include hallu- cinating extra point matches in planar regions, computing fundamental matrices directly from homographies, and applying coplanarity and other geometric constraints as part of the final bundle adjustment stage. Our experimental results indicate that the quality of the reconstruction can be significantly improved by the judicious use of geometric constraints.
|Published in||European Workshop on 3D Structure from Multiple Images of Large-Scale Environments (SMILE)|
All copyrights reserved by Springer 2007.