|
processamento de imagem e visão |
|
|
Instructors:
Lectures |
email: jsm at isr.ist.utl.pt |
|
Lab |
email: jpc at isr.ist.utl.pt |
Office hours:
JSM: Monday 15:30-17:00 and Friday 14:00-15:30
Please send an email the day before until 18:00.
Goal
This course
addresses the following question:
how to extract meaningful information from
images and video?
This is a key
question in many contexts, for example, when we wish to make robot see, when we
wish to reconstruct 3D models of the scene from multiple images, when we want
to search for similar images in a large database or when we wish to perform the
analysis of medical images to detect abnormal issues.
Some of these
problems are discussed in the course e.g., object segmentation, object
recognition and 3D reconstruction. The course also provides a hands on approach
to computer vision by involving the students in a challenging project.
Text book and material
•
[RS]
- Richard Szeliski, Computer Vision: Algorithms and
Applications, Springer, 2011 (a
pdf version is freely available at:
http://szeliski.org/Book/)
•
Slides:
available at fenix web page [link]
Each student should
attend all lessons and spend 4 hours per week, on average, preparing the Lab
and reading the textbook.
Other interesting
material
•
Computer Vision course
(University of Edimburg, Prof. Bob Fisher) http://www.inf.ed.ac.uk/teaching/courses/av/
•
SIFT slides
https://www.cs.cmu.edu/~efros/courses/AP06/presentations/0319.SIFT.ppt
Syllabus
All the topics are addressed in Szeliski
book with detail. However, we do not discuss all the models and methods
described in each chapter but s subset of them.
|
Chapter
RS |
0.
Introduction Course presentation, project
overview. |
1 |
1.
Geometry background points, lines and planes,
geometric transformations |
2 |
2.
Camera model Pin hole model, geometric
properties, image motion, extrinsic and intrinsic parameters, camera calibration |
2 |
3.
Image processing Images, linear filtering, non-linear filtering, interpolation and
pyramids. |
3 |
4.
Feature detection and matching Interest points, patches, SIFT, edges and line features, texture and
color |
4 |
5.
Global motion estimation Feature based methods, dense methods |
6 |
6.
Optical flow Optical flow equation, Horn-Schunk and Lucas-Kanade methods. |
5 |
7.
Segmentation Thresholding, region growing, snakes,
level sets. |
5 |
8.
Structure from motion Triangulation, epipolar geometry, estimation
of fundamental matrix. |
7 |
9.
Object Recognition Face and pedestrian detection. Category recognition. Bag-of-words |
14 |
Evaluation
Student grading
is based on 2 written tests (50%) and a project (50%). Minimal requirements:
9.5 in each of them.
The project is
done in groups of 3.
Important dates:
Tests: 28
Out and
12 Jan
Exam (rec): 28 Jan