GRASP Special Seminar: Tomas Pajdla, Czech Technical University in Prague, "3D Reconstruction from moving cameras - Rolling Shutter, Synchronization, and Minimal Solvers"


I will show a few results related to 3D reconstruction including: [1] 3D reconstruction with rolling shutter (RS) cameras. We will show how one can generalize the classical absolute camera pose problem, also known as P3P, to cameras with RS. [2] Analysis of degeneracies in 3D reconstruction with rolling shutter cameras. We will show that many common camera configurations, e.g. cameras with parallel readout directions, become critical and allow for a large class of ambiguities in multi-view reconstruction. [3] Simultaneous estimation of camera geometry and time shift from video sequences from multiple unsynchronized cameras based on computation of a fundamental matrix or a homography with unknown time shift between images. [4] A new method making polynomial solvers based on the action matrix method faster, by careful selection of the monomial bases. We show how to enumerate all such bases in an efficient way and that going beyond Gröbner bases leads to more efficient solvers in many cases.


Presenter's biography

Tomas Pajdla received the MSc and PhD degrees from the Czech Technical University, in Prague. He works in geometry and algebra of computer vision and robotics with emphasis on nonclassical cameras, 3D reconstruction, visual localization, place recognition and industrial vision. He contributed to introducing epipolar geometry of panoramic cameras, non-central camera models generated by linear mapping, generalized epipolar geometries, to developing solvers for minimal problems in structure from motion, solving image matching problem and to image-based localization. He co-authored works awarded prizes at OAGM 1998 and 2013, BMVC 2002 and ACCV 2014. Google Scholar: