TY  - THES
T1  - Variational 3D reconstruction from stereo image pairs and stereo sequences
A1  - Valgaerts,Levi
Y1  - 2012/01/11
N2  - This work deals with 3D reconstruction and 3D motion estimation from stereo images using variational methods that are based on dense optical flow. In the first part of the thesis, we will investigate a novel application for dense optical flow, namely the estimation of the fundamental matrix of a stereo image pair. By exploiting the high interdependency between the recovered stereo geometry and the established image correspondences, we propose a coupled refinement of the fundamental matrix and the optical flow as a second contribution, thereby improving the accuracy of both. As opposed to many existing techniques, our joint method does not solve for the camera pose and scene structure separately, but recovers them in a single optimisation step. True to our principle of joint optimisation, we further couple the dense 3D reconstruction of the scene to the estimation of its 3D motion in the final part of this thesis. This is achieved by integrating spatial and temporal information from multiple stereo pairs in a novel model for scene flow computation.
KW  - Maschinelles Sehen
KW  - Bildverarbeitung
KW  - Dreidimensionale Rekonstruktion
CY  - Saarbrücken
PB  - Universitäts- und Landesbibliothek
AD  - Postfach 151141, 66041 Saarbrücken
UR  - http://scidok.sulb.uni-saarland.de/volltexte/2012/4548
ER  -