The goal of this seminar is to provide an overview of the algorithms involved in extracting geometric information, notably depth, from two digital images.

Individual topics cover one specific aspect of such a system, ranging from the geometry and calibration of a single camera, feature detection and description, feature matching and the estimation of epipolar geometry, to rectification and disparity estimation, as well as view interpolation.

In the beginning of the term, there will be two introductory classes about projective geometry and the pinhole camera model. The final presentations will be blocked in two or three days some time at the end of the semester.

The grade is composed to 50% of a final presentation and 50% of a 6 pages seminar paper (IEEE style). Optionally, participants may opt to provide a demonstration of a working prototype after their presentation and write a shorter 3 pages seminar paper.

Credits: 5