IT-Universitetet i København
 
  Tilbage Kursusoversigt
Kursusbeskrivelse
Kursusnavn (dansk):02503 Videregående billedanalyse 
Kursusnavn (engelsk):02503 Advanced Image Analysis 
Semester:Forår 2003 
Udbydes under:cand.it., multimedieteknologi (mmt) 
Omfang i ECTS:10,00 
Kursussprog:Engelsk 
Kursushjemmeside:https://learnit.itu.dk 
Min. antal deltagere:
Forventet antal deltagere:12 
Maks. antal deltagere:20 
Formelle forudsætninger:Introduction to image analysis
Statistics 2 (principal component analysis, multivariate linear regression, discriminant analysis etc) 
Læringsmål:Understand the main topics developped in image analysis from 1980 to 2000, so that a master thesis in industrial vision, medical image analysis or computer vision theory can be started.

Hands-on experience with some very nice algorithms in MATLAB 
Fagligt indhold:1. Markov random fields: Dynamics and use for segmentation (Ising and Potts model)
2. Snakes
3. Statistical shape modelling using minimum description length (MDL)
4. Active shape model
5. Active appearance model
6. Level sets for segmentation
7. Fast marching method for segmentation
8. Gabor filters for texture analysis
9. Mathematical morphology: greyscale morphology
10. Multiple view geometry (Henrik Aanæs)

At present the teacher is Hans Henrik Thodberg, who has worked many years in industry.
The spirit of the course is to present useful algorithms rather than academic excursions.


If you are not good at MATLAB, you will be. 
Læringsaktiviteter:

13 double lectures
7 MATLAB exercises
1 large project (of your own choice) to be presented as a poster at the annual vision meeting at IMM \"visiondagen\" in June 

Eksamensform og -beskrivelse:X. experimental examination form (7-scale; external exam), Bestået/ikke bestået, Intern censur

All exercises must be completed successfully.  

Litteratur udover forskningsartikler:Book chapters, articles, reviews
see last year\'s homepage www.imm.dtu.dk/courses/02503, with links to downloads, references etc.
The powerpoint lectures are also posted here