This is the
course homepage for 02503 advanced image analysis running at IMM, DTU every
spring semester (entry
in DTU course catalogue). In general the course runs on Wednesdays,
starting at
The course
is a 10 point course with a graded on the 7 point scale. The course runs once a
week for 13 weeks. Eleven of these times have lectures in the morning ( starting at 9 O’clock) and followed by
exercises, the last two times consists only of exercises, aimed at illustrating
the systems approach to image analysis and at exploring the interlink between
the different subjects covered. Note that 10 points correspond to ca. 18 hours
of work pr. week for the average student, much of this time is expected to be
used on solving the exercises, also outside the the schedualed time slot where
the TA will be present.
Lectures will take place in room 033 of
building 322 from 9.00. The exercises will take place in databar 113 off
building 305.
The course
material consists of the notes, book chapters and articles listed in the
overview below and/or on CampusNet.
Electronic material is obtained by a click of the mouse. Otherwise you will get
photocopies at the lecture at least one week in advance. Clicking the lecture
title sometimes brings you to the pdf files of the lecture slides, also try CampusNet.
The exam is
oral and based on the curriculum (most of the course material) and the
completed exercises. Technically the exercises are not mandatory, i.e. it will
not be checked if you have completed them, you will just be held accountable
for the results you have obtained at the oral exam. The first exercise “4 small exercises” is not part of the curriculum.
The students
, however, have the offer of having reports/journals of your exercises being
corrected, if you hand them in no later than two weeks after the exercise has
been posed. Some times that deadline is extended, until enough reports have
been handed in, but this is not a guarantee.
The oral
exam proceeds by randomly choosing one of the exercises, which the proceeding
questioning will take its outset in, but not be limited to. The student is not
expected to give a presentation, just be able to account for the curriculum and
the completed exercises. The duration of a questioning will be approximately 20
minutes.
SUBJECT TO CHANGE!
Lecture Number |
Morning Lecture |
Afternoon Exercise |
1. 2/2 |
Local Differential Image Properties. Reading material: Scale-space theory: A framework for handling
image structures at multiple scales, by Tony
Lindeberg |
Exercise to get you up to speed on the
prerequisites. Not curriculum for oral exam. Data: Suplemetary reading material: Knut Conradsen: An
Introduction to Statistics, Vol 2 (excerpts) The Multivariate Normal Dist.
Discriminant Analysis Principal Components |
2. 9/2 |
Markov random fields Reading Material: |
Data: |
3. 16/2 |
Texture – Lecture given by Anders Dahl Reading
Material: |
Exercise in Markov Random Fields Data and relevant functions: Compiled Mex file for windows users: |
4. 23/2 |
Object recognition – Lecture given by Anders Dahl & Henrik
Aanæs Reading
Material: ·
Part 2 of my lecture notes,
with special emphasis on SIFT features, ·
Sivic and Zisserman (2003): Video
Google: a text retrieval approach to object matching in videos. ICCV 2003. [PDF] ·
“Robust real-Time Face Detection”
Viola and Jones IJCV 2004 [PDF] |
Impainting – to appear Responsible ABD. |
5. 2/3 |
Large Exercice on: Multi Label Texture Segmentation - to appear Responsible Anders Dahl & Henrik Aanæs NB: For the students
needing a refreshment of the course prerequisites in statistics, a short overview will be given
at 9.15. |
|
6. 9/3 |
Active Shape Models - Lecture Given by Rasmus Larsen Reading material: Statistical Models of Appearance
for computer vision. By T.F. Cootes and C.J.Taylor.
Focus on sec 4.1 to 4.5 and dec. 5.1 and 5.2. |
|
7. 16/3 |
Orthogonal Image Transformations: - Lecture given by Allan Aasbjerg Nielsen Reading
material: v Allan Aasbjerg
Nielsen (2007). The Regularized
Iteratively Reweighted MAD Method for Change Detection in Multi- and
Hyperspectral Data. IEEE Transactions on Image Processing 16(2), 463-478, under http://www.imm.dtu.dk/~aa/ v Allan A. Nielsen
and Morton J. Canty (2008). Kernel principal
component analysis for change detection. SPIE vol. 7109, Europe Remote
Sensing Conference, Cardiff, Great Britain, 15-18 September
2008, under http://www.imm.dtu.dk/~aa/ v and for the really keen: v Morton J. Canty and Allan A.
Nielsen (2008). Automatic
Radiometric Normalization of Multitemporal Satellite Imagery with the
Iteratively Re-weighted MAD Transformation. Remote Sensing of Environment 112(3), 1025-1036, under http://www.imm.dtu.dk/~aa/ v Morton J. Canty Allan A. Nielsen
and Michael Schmidt (2004). Automatic
radiometric normalization of multitemporal satellite imagery. Remote Sensing of Environment 91(3-4), 441-451, under http://www.imm.dtu.dk/~aa/ |
Data and relevant functions: cars.mat |
8. 23/3 |
Large Exercise on: Face Recognition - to appear Responsible Rasmus Larsen |
|
9. 30/3 |
Active Sensors & 3D Reconstruction Reading material: ·
Chapter on Active 3D Sensors from
Henrik Aanæs’s Notes on Computer Vision, ·
Chapter 14 from “Computational Geometry
Processing” by Anton et al. |
|
10. 6/4 |
Stereo Reading material: ·
G. Van Meerbergen, M. Vergauwen, M. Pollefeys, L. Van
Gool. A Hierarchical Symmetric Stereo Algorithm Using Dynamic Programming,
International Journal on Computer Vision 47(1/2/3): 275-285, 2002. [pdf] |
Structured Light Scanner- to appear Responsible Henrik Aanæs |
11. 13/4 |
3D Volumetric Reconstruction - Lecture given by Rasmus Paulsen |
Two View Stereo via MRF Responsible Henrik Aanæs |
Easter Holliday |
||
12. 27/4 |
Large Scale Reconstruction and Mapping |
Exercise in 3D reconstruction - to be defined |
13. 4/5 |
Complex Distributions in
Image Analysis -
Lecture Given by
Knut Conradsen |
Structure and Motion
Exercise Responsible |
Last
updated: 30/03/2011 08:23 by