Title

• Graph Signal Processing
• Traitement de signaux sur graphes

Identification

• ANR-14-CE27-0001

Duration

• 42 Months (May 2015 - October 2018)

Coordinators

• Project and partner 1: Olivier Lézoray, GREYC, Université de Caen Basse-Normandie
• Partner 2: Pierre Borgnat, Laboratoire de Physique, Ecole Normale Supérieure de Lyon
• Partner 3: Sophie Achard, GIPSA-lab, CNRS
• Partner 4: Emilie Chouzenoux, LIGM, Université Paris Est Marne La Vallée

Context and purposes

An increasing number of application areas require the processing of massive datasets. These data can often be represented by graphs in order to encode complex interactions. When data vectors are associated with graph vertices or edges, a so-called graph signal is obtained. Processing such graph signals includes several open challenges because of the nature of the involved information. Indeed the combination of graph theory and signal / image processing methodologies has not been studied as much as it should. Examples of such graphs signal are shown below.

On the one hand, new development are needed in order to allow classical signal processing methods to work on irregular grids and non Euclidean spaces. On the other hand, considering the significant success of classical signal processing tools, it appears highly desirable to generalize their use to graph signals. The GRAPHSIP project aims at developing a set of advanced methods and algorithms for the processing of graph signals: in particular multi-scale transforms and solutions of variational problems on graphs. The major outcomes of this project are expected to lead to significant breakthroughs for graph data processing. The project will also focus on two novel applications on instances of graph signals: brain networks and 3D color point clouds. They will exemplify and illustrate the proposed methodological advances on emerging applications.

Work-Packages