bild
School of
Electrical Engineering
and Computer Science

Image processing and computer vision, 6 credits

Please note that this course was last given in 2009.

New course info (from 2010): bildat10.

Latest course info (2009): bildat09.

Humans use vision as the primary sensor to gather the information about the environment. Computer vision aims at implementing similar capabilities on artificial systems (computers, robots). It deals with development of algorithms and computational models that can automatically process and extract data from digital images or sequences.

This course dives and introduction to the wide research field of computer vision and image analysis, image processing and image compression. All of the above reasearch areas are nowadays in a rapid stage of development and it is expected that their importance will be more and more significant. As some of the application areas we can mention robotic systems with visual capabilities, medical image processing for smart medical/operating rooms, image based material quality inspection, three dimensional modelling of space, man-machine interaction, advanced image compression and processing of satelite images.

At the beginning of the course, we will start with some basic image operators so to detect and extract different types of low level image cues. This is mostly related to the image processing part where we refer to signal processing techniques in two dimensions. Some examples are grey-level tarnsformations, image filtering and detection of features such as edges and corners. These methods and of general nature and serve as components in most of the complex vision systemsand applications mentioned above.

Related to the computer vision part of the course, we will study different methods for extracting the threedimensional information from images and use it for reconstruction of the world observed by a cemera. Specific problems that we will concentrate on are texture, stereo and motion cues. As an example, we will show how some simple assumptions about the world can be used in a combination with a thorough mathematical modeling for reconstruction. In addition, we will show how we can define some invariant properties that do not change under perspective projection.

As a part of a course, examples of how a robot that moves in an environment can use computer vision to avoid obstacles by estimating so called time-to-contact will be given. In addition, methods for object detection and recognition will be studied.

In summary, computer vision is an interdisciplinary research field and it strongly relates to biological and human visual systems. During the course, we will also spend some time on presenting and illustrating human visual system and human perception.

This course is a part of specialization Autonomous systems and it is given by the Computational Vision and Active Perception Laboratory . This group, shortly named CVAP, is comprised of approximately 30 persons taht pursue reaserch on both basic theories and modeling of complex artificial vision systems with application in the areas of robotics, medical image processing and visual perception in general.

PREREQUISITIES

We suggest a sound basic knowledge in applied math and computer science, especially linear algebra and numerical analysis. Additional courses related to digital signal processing are recommended.

Text in the study handbook in swedish or in english.

Copyright © Published by: Danica Kragic <danik@nada.kth.se>
Updated 2009-12-17