Overview of the Research Activities in the IMAGE Team

The GREYC Image team is carrying out research around three theoretical themes: "Geometrical approaches for image processing", "Estimation, Detection and Pattern recognition" and "Knowledge Engineering for Image Processing". The team is composed with researchers from various fields (physics, mathematics, compute sciences, artificial intelligence). This wide range of competences allows to handle the problem of image processing and understanding from several complementary viewpoints. Browse the different research activities of the Image Team with the menu on the left.

The research undertaken in this group aims at studying the theoretical and methodological aspects related to image acquisition, detection, estimation and pattern recognition. These skills allow us to tackle problems of a complete application covering all the image chain going from acquisition and its induced problems (degradations, noise, etc) until decision-making. The fundamental part of our work is carried out by bringing together tools from three theoretical frameworks : harmonic analysis (wavelets and beyond), statistical approach, variational approach and PDE-based image processing. These frameworks offer all the necessary ingredients for the development of original methods to process and analyse multidimensional images such as non-linear restoration, non-linear reconstruction and interpolation, segmentation and tracking, statistical pattern training and recognition. Applications of these methodologies are focusing, without restriction, on multimodal medical imaging (e.g. functional anatomical MRI, diffusion MRI, echography, EEG), image and signal content security, biometry and multi-media applications. The members of this group have yet developed scientific collaborations with academic collaborators (such as Cambridge university, Brisbane university, Sophia Antipolis university, INRIA Sophia Antipolis, CEA Saclay, Cyceron) and also with industrial partners (Philips, France Telecom R&D, Institut Francais du Pétrole, TCI).

The theme "Knowledge Ingeneering for Image Processing" aims at studying image processing applications development. The motivation is not only to help the image processing specialists to develop complex softwares, but also to build a patrimony of skill knowledge.

The developement of an image processing is decribed as a complex activity. Even if effective processings rely on numerical theories, the conception of a software solution is not a purely numerical problem. We propose to link together the requirement analysis to the conception of a suitable solution. The paradigm relies on the supervision of a library of operators where the conception of a solution consists in selecting convenient operators from the library, tuning their parameters and linking them to compose graph of operators.

Our short term objective is to master the image analysis know-how and to capitalise the underlying knowledge that is involved, through the study of effective developments: the workbench makes the development more easier and the development increases the patrimony of knowledge.

Our long-term objective is the design of a knowledge-based system for the automation of the generation of image processing programs from a request formulated by a user unskilled in image processing.