Lyon Research Centre for Images and Intelligent Information

 

LIRIS (Laboratoire d'informatique en image et systèmes d'information - Imaging and Information Technology Laboratory) was formed at the beginning of 2003 from the reorganisation of several Lyon-based research laboratories, namely LIGIM, LISI and RFV and other entities operating individually in the domain of Information and Communication Technology and Sciences. LIRIS is affiliated to the CNRS under reference FRE 2672 and covers two main research domains, Digital Imaging and Information Systems, subdivided into the following four scientific fields:

• Field 1 - Data, Document and Knowledge Bases
• Field 2 - Images and Video: Segmentation and Extraction of Information
• Field 3 - Modelling and Enhanced Reality 
• Field 4 - Communicating Information Systems
  

with two cross-cutting objectives:

• Objective A - Digital Document Software Integration Platform, in conjunction with the ISDN (Institut des sciences du document numérique - Institute of Digital Document Sciences).
• Objective B - Software Integration Platform: Distributed Medical Multimedia File, under the scope of Health Engineering.
  

LIRIS comprises around 150 individuals including nearly 70 research lecturers in four teaching centres in Lyon: INSA de Lyon, the Université Claude Bernard Lyon 1, the École Centrale de Lyon and the Université Lumière Lyon 2 along with other sites in La Doua, Ecully and Bron.

The Lyon 2 LIRIS team concentrates on the various aspects of modelling, analysis and processing of digital images. The Lyon 2 researchers are principally attached to Laboratory Field 3 (Geometric Modelling) in close co-operation with Field 2 (Images) and Field 4 (Communicating Information Systems). They are also heavily involved in Laboratory Cross-Cutting Objective B (Health Engineering).

For several years now, the LIRIS-Lyon 2 researchers have been major participants in both the French and International discrete geometry communities. This domain of research studies the arithmetic and geometric properties of sets of points in digital images. It finds applications which recognise or describe shapes and compress data and demonstrates the great advantage of whole number calculation which does away with the approximations and interpolations inherent in floating point calculation.

The principal field of application for our research is medical imaging. We have collaborated on numerous occasions with the Centre de lutte contre le cancer Léon Bérard (Léon Bérard Cancer Research Centre). Part of our team is currently on secondment to the latter's Radiotherapy department and is collaborating on improving an irradiation protocol using image analysis, processing and synthesis techniques. We are also working on a breast cancer diagnosis assistance system based on the automatic analysis of mammography scans.

Processing very large image databases requires computing power far in excess of that available to centralised systems alone. The medical imaging domain is further complicated by the high level of distribution of the information and its often confidential nature. For several years now, a number of research teams around the world have been concentrating their efforts on the possibility of using the latent power of networked computers and the mechanisms which would enable these resources to be pooled in order to form a computing grid. We are involved in several projects aimed at studying how these grids can be applied to the processing of medical images.

Since the end of 2003, LIRIS has been a partner of Foxstream, a newly created company supported by Créalys, the Inter-University Incubator.  Foxstream specialises in remote video surveillance and parts of its activities are based on the transfer of technologies produced by LIRIS's image processing and analysis research. In the security domain, an ever-increasing number of video cameras are being installed in sensitive sites. Foxstream provides integrated solutions for intelligent surveillance of such sites which are capable of recognizing certain predefined behaviours so that they do not set off a false alarm. For example, a pedestrian entering a car park could be categorised as a suspicious event whereas a vehicle entering the same car park could be categorised as a normal event. Similarly, systems must be capable of distinguishing between a patient having a fall in their home and a normal movement picked up by the camera.