Image processing and computer vision

lelec2885  2019-2020  Louvain-la-Neuve

Image processing and computer vision
Note from June 29, 2020
Although we do not yet know how long the social distancing related to the Covid-19 pandemic will last, and regardless of the changes that had to be made in the evaluation of the June 2020 session in relation to what is provided for in this learning unit description, new learnig unit evaluation methods may still be adopted by the teachers; details of these methods have been - or will be - communicated to the students by the teachers, as soon as possible.
5 credits
30.0 h + 30.0 h
De Vleeschouwer Christophe (coordinator); Jacques Laurent;
Main themes
This course is part of the ELEC/EPL program in "information and signal processing". The main objective of LELEC2885is to introduce all the concepts needed to understand the "image" signals, from their acquisition until their processing, through the important questions of signal representation and approximation occuring during data transmission or interpretation.

At the end of this learning unit, the student is able to :

1 With respect to the AA referring system defined for the Master in Electrical Engineering, the course contributes to the develoopment, mastery and assessment of the following skills :
  • AA1.1, AA1.2
  • AA3.1, AA3.3
  • AA5.5, AA5.6
b.   At the end of this course, the student will be able to:
1. Handle techniques of representation and approximation of images in order to extract their meaningful components with respect to a particular application, for example, in the fields of data transmission or interpretation;
2. Apply linear and non-linear filtering operations (e.g., morphological) to isolate certain frequency components or to cancel particular noises;
3. Detect structures of interest in an image, such as contours, key features, etc..
4. Segment an image into regions of homogeneous characteristics, targeting a semantic interpretation of the image content;
5. Restore images corrupted a noise or a blurring;
6. Understand the basic principles of inverse problem solving in imaging and in compressed sensing;
7. Manage image databases using detection tools or classification;
8. Detect and track one or more object(s) of interest in video streams, in biomedical applications or for 3-D scene interpretation;
9. Compress image signals considering their visual perception and their accessability in the compressed signal representation;
10. Provide a solution to complex problems involving image processing, such as quality control, visiosurveillance, multimodal human-machine interfaces, and image compression.

The contribution of this Teaching Unit to the development and command of the skills and learning outcomes of the programme(s) can be accessed at the end of this sheet, in the section entitled “Programmes/courses offering this Teaching Unit”.
  • Image representation: Pixels, Fourier and Multiscale Transforms.
  • The wavelet transform.
  • The sparsity principle and applications: from orthonormal bases to redundant systems.
  • Human visual system and salient image features.
  • Image classification and deep learning introduction.
  • Basic tools of image analysis: mathematical morphology and relatives.
  • Image segmentation, (spectral) clustering, watershed and level sets
  • An introduction to computational imaging
  • Detection-based (multi-) object tracking: detect-before-track
  • Recursive visual object tracking: track-before-detect
  • Principles of stereo vision
  • From entropy coding to image compression
  • Video compression, and sparse approximation coding
Teaching methods
The course is organized around a series of lectures, each dealing with a specific problem commonly encountered in the field of image processing. Each lesson introduces a selection of the main solutions found in the literature and/or the industry to solve the problem of interest, and a list of references is provided for each covered topic.
To complement the lectures, the student is also asked to read and criticize a number of scientific publications. The goal is to allow him/her to deal with a subject in depth, but also and especially to draw his/her attention to the way a scientific paper is built.

In addition to the theoretical classes, numerical exercise sessions under Python are organized in a computer room. Students are asked to program different algorithms associated with a consistent sub-selection of the techniques taught. They use existing Python libraries for this purpose. Learning is provided by problem solving, based on real or synthetic images/signals, sometimes associated with external databases.
The course is given in the classroom exclusively.
Evaluation methods
The evaluation includes three components :
  • An oral examination: Scheduled in January, this test evaluates individually the students on their understanding of the concepts and methods taught during the theoretical courses.
  • An evaluation of the Python numerical exercises: students are evaluated on a computer (in session or out of session) based on problems similar to those presented during the year.
  • A critical analysis of 3 scientific papers in the field: This helps the student to develop his ability to analyze the advantages and the weaknesses of a scientific work, considering both its content and its general structure. Each student provides a report (1 page max per article) by December.
These three components are weighted as 50%, 30% and 20% of the final grade, respectively. It is required to pass the oral exam. In case of exam failure, only the exam note will be taken into account.
Other information
This course assumes that the basics of signal processing, such as taught in the course "signals and systems" (LFSAB1106) or "digital signal processing" (LELEC2900), are known.
Support de cours :
Transparents, articles tutoriaux et parties de code Python.
Les documents du cours sont disponibles sur Moodle
Lectures conseillées :
Durant l'année, l'étudiant doit lire 3 articles sélectionnés dans une liste d’articles distribués sur le site Moodle du cours.
Course materials:
Slides, tutorials and parts of Python code.
Course documents are available on Moodle
Recommended reading:
During the year, each student must read 3 articles selected from a list of articles distributed on the Moodle site of the course.
Faculty or entity

Programmes / formations proposant cette unité d'enseignement (UE)

Title of the programme
Master [120] in Data Science Engineering

Master [120] in Biomedical Engineering

Master [120] in Computer Science and Engineering

Master [120] in Mathematical Engineering

Master [120] in Computer Science

Master [120] in Electrical Engineering

Master [120] in Data Science: Information Technology