After successful completion of the course, students are able to name the basic processing steps of stereo vision and to point out solutions for potential problems in this context. Students are able to identify local, global, and semi-global stereo matching processes, describe how they work and assess their relevance/applicability in a set scenario. Furthermore, students are able to evaluate stereo vision algorithms in regard to the state of the art and to compare them to referential results of relevant benchmarks.
The main topic of the lecture is the solution of the so-called stereo correspondence problem, in which a corresponding pixel in the right image has to be found for each pixel in the left image. This task represents one of the central challenges in the field of Computer Vision. The first lectures will feature an introduction to local stereo methods (block-matching). In the following, there will be a focus on global/ semi-global approaches (dynamic programming, belief propagation and graph-cuts). Additionally, the evaluation of stereo matching algorithms with the help of relevant benchmarks will be discussed. The practical course will feature the implementation and evaluation of a simple stereo-matching algorithm.
The lectures and the lecture slides (which will be made available on TUWEL) are supplemented by suitable video material, demos, additional reading and weblinks, in order to illustrate the lecture content and current research questions.
The lectures will be held from 10AM - 12PM every Wednesday between April 17th and June 12th 2024. The last two dates are planned as alternative lecture dates in case one of the other lecture dates has to be postponed. The preliminary meeting will take place on March 20th 2024 from 10AM - 12PM in the Seminarraum FAV 01 B (Seminarraum 187/2).
ECTS Breakdown: 3 ECTS = 75h14h Lecture30h Exercise implementation5h Documentation of exercise results, discussion of results (Abgabegespräche)24h Exam preparation2h Exam---75h
Please note that this course will not be offered every year; thus, the next course will be offered in the summer term 2026.
The exam serves to assess how well the processes and algorithms presented in the lecture have been understood. You will be asked to give your answers as written explanations, formulas, sketches, pseudo-code etc. The final grade for the practical course consists of the grades you receive for training exercises that you have to hand in and a final conversation.
The final grade will be calculated from your exam results and the marks you receive in the practical part of this lecture.
Registration in TISS. Participation in the pre-lecture meeting (Vorbesprechung) is compulsory.
The course material will be made available via TUWEL.
The course requires basic knowledge of image processing as taught in relevant courses of the bachelor programme Media Informatics and Visual Computing (e.g., VU 186.822).