Des otolithes aux satellites: méthodes et applications du traitement du signal et des images pour l'observation de l'océan

This document presents a synthesis of my research activities over the last 10 years initially within the Ifremer-IRD schlerochronology laboratory and the Fisheries Science and Technology department at Ifremer, and then at the Signal & Communication Department of Telecom Bretagne. My research activities were undertaken at the interface of Information Science and Technology and Oceanography. In the framework of interdisciplinary approaches, my research work addressed the development of new image and signal processing tools and methods with a view to (1) providing new representations of the obsvered scenes or processes, (2) exploiting these representations to infer or reconstruct patterns of interest for the considered thematic objectives. Three thematic issues were addressed in the field of the remote sensing of the ocean (in a broad sense): initially the otoliths as recorders of the individual life traits of fish and the acoustic sensing of the seabed and of the pelagic ecosystem, and more recently the satellite-based remote sensing of the ocean surface. This research involved generic methodological developments in the field of information processing and pattern recognition such as the analysis of the geometry of multivariate signals (including shape analysis), texture analysis and recognition, missing data interpolation, object and scene reocgnition using various methodological frameworks (probabilistic models, Bayesian and variational inference, statistical learning,...). From this expertise I wish to explore the potential, widely unexplored, of the existing databases of multimodal and multiscale observations of the ocean for the characterization and the modelling of the dynamics of marine ecosystems. These thematic issues involve key information processing challenges which will be at the core of the multidisciplinary research I will undertake in the coming years.