« The new generation of high-resolution sensors makes it possible to characterise objects with ever finer resolution. Processing these textured images requires the development of new multivariate stochastic models adapted to the specific characteristics of the signal being analysed, such as spatial, temporal and/or multi-modal dependence. While the processing of one-dimensional signals is now the subject of numerous theoretical developments that have been perfectly mastered by the community, the extension to multi-dimensional signals still comes up against numerous methodological problems. Multivariate approaches are therefore emerging as a way of analysing and extracting information from these images. My research activities at the IMS laboratory focus on two main areas. The first concerns the development of multivariate stochastic models and similarity measures derived from information geometry. The second focuses on signal and image processing on manifold, in particular on the space of covariance matrices. This research lies at the interface between advanced image processing methods and agri-environmental applications such as weed detection and leaf symptom recognition. » – Lionel Bombrun.
