Supervisory authorities

CNRS Université Jean Monnet Institut d'Optique Graduate School


Home page > Research groups > Image processing

Image processing

Group leader: Christophe Ducottet

In our so-called information society, the digital image is widely used in many areas such as photography, web, video, television, cinema, medical or biological imaging, satellite imagery, ... How to control the processing, retrieval and management of this wonderful means of communication and knowledge production?

The scientific challenge of our group is to provide techniques, algorithms and models to better harness the potential of images and videos in information technologies.

Examples of scientific bottlenecks we address are:

- Automatic extraction of semantic content of images and videos for object recognition or information retrieval tasks.
- Multimodal analysis and information retrieval by exploiting the synergy between the various media, including text, color and psycho-visual information.
- Tomographic reconstruction suited for high resolution and low-dose radiation in medical imaging.
- Time resolved 3D information reconstruction in digital holography for fluid mechanics and biological imaging.

Principal research directions

1. Image reconstruction

Leaders: Rolf Clackdoyle and Corinne Fournier

Considering advanced imaging systems as tomography or holography, the goal is to reconstruct information about imaged objects. Inverse-problem based approaches are developed using physical models of objects and image formation. Both theoretical and more practical issues are addressed either to study the limits of imaging systems (conditions for reconstruction, resolution ...) or to develop efficient reconstruction algorithms.

2. Image content extraction and representation

Leader: Alain Trémeau

The problem of visual content interpretation is addressed. What information to extract from images and how to index it efficiently for applications as image and video retrieval, document classification or object recognition? The originality of our approach is to specifically consider color information and to use data mining techniques and learning from structured data (strings, trees and graphs) in interaction with researchers of the machine learning team.

This section doesn't contain any article.