Abstract:

Epipolar geometry estimation from image pairs is a basic problem in computer vision. It is used as a component of many important applications such as vision based robot navigation, structure from motion and other multiple view geometry applications. Even though successful algorithms have been proposed to address this problem, it still remains an active field of research.

In this talk I'll present:

1. Most successful epipolar geometry estimation algorithms.

2. Several reasons why input given to these algorithms may be challenging.

3. Preprocessing method improving these algorithms results considerably, so they are now able to succeed on hard cases for which they failed before.

Registration

*No prior Computer Vision/ Machine Learning knowledge is needed.