Visual Content Analysis

Catalog of Télécom SudParis courses

Code

IGSF IMA 4509

Level

M1

Graduate

Graduate

Semester

Spring

Domain

Image

Program

Programme Ingénieur

Language

Anglais/English

ECTS Credits

4

Class hours

45

Workload

90

Program Manager(s)

Department

  • Advanced Research and Techniques for Multidimensional Imaging Systems

Educational team

Organisation

Cours/TD/TP/projet/examen : 36/0/9/0

Learning objectives

At the end of this teaching unit, students will be able

- To model an inverse imaging problem under a mathematically well-posed form via a deterministic or stochastic regularization approach, and to solve it using companion optimization methods.

- To specify & use core techniques for solving basic low-level image & video analysis problems, including restoration, multiscale description, point of interest detection, segmentation, calibration & 3D scene reconstruction, motion analysis & tracking.

- To design & conduct an experimental performance study of an image understanding technique, relying on dedicated software tools such as the industry-standard Matlab(R) platform or the open-source OpenCV library.

- To write a synthetic and reasoned critical analysis of a scientific article dealing with an image understanding or computer vision issue.

CDIO Skills

  • 1.3 - Advanced engineering fundamental knowledge, methods and tools
  • 2.1 - Analytical reasoning and problem solving
  • 2.2 - Experimentation, investigation and knowledge discovery
  • 3.2.3 - Written Communication
  • 4.7.3 - Thinking Creatively and Imagining Possibilites (which builds on and expands Creative Thinking 2.4.3)

Prerequisites

None

Keywords

Visual feature extraction; denoising, enhancement & restoration; segmentation & grouping; motion estimation & tracking; shape analysis.

Content

> Digital image & video analysis: imaging chain, paradigms and models

> Still image analysis

- Advanced image filtering & feature detection
• Interest point/line detection: corners, ridges and valleys
• Non-local image filtering: bilateral filters, Non-Local means, patch-based image denoising

- Gray-level mathematical morphology
• basic Euclidean & geodesic operators
• morphological filtering
• morphological segmentation

- Inverse problems in image analysis: deterministic & stochastic regularization

-Deterministic image segmentation: variational methods & graph cuts
• Active contours: parameterized models, level set methods
• Active regions: the Mumford-Shah model, deterministic & statistical region competition, multi-feature variational segmentation

- Scale-space & PDE image filtering

- Bayesian methods & Markov Random Fields

> Multi-view image analysis
- Calibration
- Stereovision & multi-view 3D scene/object reconstruction

> Video analysis
- Motion measurement & optical flow estimation
- Spatio-temporal segmentation & object tracking

Evaluation

The assessment pattern involves 3 components: continuous evaluation via homework on selected topics (”coursework”) (CW), lab assignments (L), and a two-student group written final exam (E). The final grade is a weighted average of individual component grades. The 2nd session consists of a study with an oral defense (O).
-1st session = Weighted Average (CW, L, E)) (S1)
-2nd session = O (S2)
-Final grade = Max (SE1, SE2)