Computer Vision in 3D

Examiner and lecturer:

J. Bigun                 


Litterature:

[1] J. Bigun, "Vision with direction",  Springer, (2006).

[2] Laboratory exercises in Matlab: 

Optical Flow by DFD and Correlation,
Optical Flow LK,
Calibration,
Depth by Stereo,
Epipolar Geometry.

Course Contents:

The course consists in two parts: Fundamentals of computer vision in 3D and Scientific paper. Examinations of the two parts are independent.

Fundamentals of Computer Vision in 3D

This part consists in two concepts, theory and labs (also called practice). Both are examined in writing. A grade of 3,4 or 5 represents Pass.

Theory supports student learning in terms of understanding of tools necessary to construct computational methods. Practice is meant to give skills illustrating the theory. Evidently, not every detail of theory is possible to illustrate via labs. Likewise, not every detail of practice is explained in terms of theory, as some practice contents are expected to be supported by your prior knowledge, e.g. basic programming skills and implementation of basic math theory in programs. Below is the summary of theory and how this relates to practice as well as sections of the course book.

Motion analysis (Ch. 12)
Brightness constancy, real motion, perceived motion, Optical flow (Ch. 12.12, 12.4, 12.6)
Correlation based methods Slides Bigun_VD2006_Ch12.12,10,4,6 (Ch. 12.10 LAB 1)
Gradient (2D) based methods, Slides Bigun_VD2006_Ch12.12,9,4,6. (Ch. 12.9 LAB 2)

Perspective mapping, calibration, stereo, 3-D reconstruction, and epipolar geometry (Ch. 13)
Perspective mapping, World basis, Camera basis, Camera calibration, Slides Bigun_VD2006_Ch13.1-3. (Ch. 13.1, 13.2, 13.3 & LAB 3)
Geometry of stereo cameras, Depth by stereo    (Ch. 13.4, 13.8 & LAB 4)
Epipolar Geometry: Facilitation of searches for correspondence    (Ch. 13.5, 13.6, 13.8 & LAB 5)


Laboratory exercises material (se Litterature above) contain normally a pdf document, images and matlab code which can be incomplete or complete. The pdf document will guide you through the lab duties and ask for explanations, or completitions of matlab codes when they are not. Questions in pdf documets ARE NOT THE SAME, as the questions in examination.

Scientific paper

As part of the course, the student will be asked to study a scientific paper, which will be evaluated by a written exam to result either in Pass or Fail.