Computer Vision [CV]
Dates & Rooms:
Monday, 10:15 - 11:45; Room: 02.133-113
Tuesday, 12:15 - 13:45; Room: 02.133-113
The exams will be on the following dates:
Monday 30. July 2012
Thursday 30. August 2012
Tuesday 25. September 2012
If you register in MeinCampus (for some students, e.g. SAOT, this is not possible), your grade will be directly entered in the system, otherwise you will have to pick up a Graded Certificate from us.
Either way, you must reserve a time slot for an exam. To do so, either email one of the secrataries, Kristina Müller or Iris Koppe , or personally visit the secretaries office at 09.138.
Computer Vision involves the (semi-) automatic extraction of
information from images. The image data itself can take many forms:
color or black-and-white images, video sequences, multiple cameras,
data from medical scanners, etc. The information that should be
extracted can also vary depending on the application: locating an
object in an image (image database search), precisely measuring the
dimensions of an object (quality control), following a moving item
(surveillance), identifying letters and numbers (optical character
recognition), estimating the position and orientation of a specific
object (robot arm guidance), etc. As a result, the field of computer
vision covers a wide variety of topics, which may sometimes, at first
glance, seem unrelated.
This course provides an introduction to the field of Computer Vision,
focusing on the underlying algorithmic, geometric and optic issues. It
starts with a description of image formation, including geometric,
optic and electronic aspects of the image formation process. Lower
level algorithms are then presented on the extraction of different
types of image features (edge detection, texture, color,
multi-resolution analysis, Hough transform, deformable contours). The
course will also cover topics associated with extracting information
from multiple images (stereo, motion). The last set of topics will
cover higher level analysis like grouping, and classification with
examples on image retrieval and face detection.
The schedule of the lectures is subject to change.
It is regularly updated to more accurately reflect what has been covered so far in the lectures. The updated version can be found here.
The slides will become available as soon as possible, but usually shortly after the lecture.
Slides from last year can be found here.
|Introduction:||A brief introduction to the various topics of computer vision, course motivation and guidelines.|
|Image Formation:||Lens, radiometry, geometric optics, coordinate systems, projection.|
|Cameras:||Digital image capture: from image irradiance to pixel values.|
|Smoothing:||Sensor noise and methods for reducing image noise, convolution.|
|Edge Detection:||Gradient-based edge detection, Canny edge detector, Laplacian of Gaussian, Gaussian pyramid, Laplacian pyramid.|
|Texture:||Texture recognition, oriented filters, texture synthesis, shape from texture.|
|Color:||The physics of color, trichromacy, color perception, color spaces, example applications.|
|Hough Transform:||Line detection, circle detection, ellipse detection, HT for arbitrary shapes.|
|Deformable Contours:||Active contours, energy functional, greedy minimization, implementation adaptations.|
Basic binocular stereo setup, disparity, triangulation, correspondence problem
|Structured Light:||Structured light setup, triangulation, binary coding, Kinect sensor|
|Multiview Geometry:||Epipolar geometry, epipolar constraint, eight-point algorithm|
|Motion Analysis:||Background subtraction, optic flow, motion field, optic flow computation|
|Kalman Filtering:||Predictive motion analysis, dynamic system under observation, Kalman filter formulation, extended Kalman filter.|
|Particle Filtering:||Markovian dynamic systems, Bayesian estimation, particle filters, marginalized particle filters. |
|SIFT:||Scale Invariant Feature Transform, keypoint detector, SIFT feature vector construction, matching SIFT vectors.|
1. If you want to know the formal justification behind the Non-Maximal Suppression and the Hysteresis Thresholding in Canny's edge detector, read the original paper:
John Canny. "A Computational Approach to Edge Detection." IEEE Transactions on Pattern Analysis and Machine Intelligence. Vol. 8, No. 8, pp. 679 - 698. 1986.
2. A very nice review article on the various stereo correspondence articles is:
Daniel Scharstein and Richard Szeliski. "A Taxonomy and Evaluation of Dense Two-Frame Stereo Correspondence Algorithms." International Journal of Computer Vision, Vol. 47, pp. 7 - 42. 2002.
3. For more information on Covariance Matrices follow this link.
4. A detailed tutorial on Principal Component Analysis can be found here.