Laure Blanc-Féraud is CNRS Research Director at I3S lab (University Côte d’Azur/ CNRS), in the Morpheme team, shared with Inria and the Institute of Biology Valrose (IBV) in Sophia Antipolis in France. Her research topic concerns image processing, mainly inverse problems, using PDE and calculus of variation, under smooth, non smooth and l0-sparse constraints. She studies minimization problems using duality, convex and non convex optimization, Γ-convergence. She is also devoping bayesian modeling for model parameter estimation. Since 2011, she focuses her activity on 3D microscopy imaging in biology, mainly in super-resolution technics and extra cellular matrix characterization.
She is member of the IEEE Technical Committee BISP, has been associate editor of the journals SIAM Imaging Science (13-18) and Traitement du Signal, director of the French national Research group GDR ISIS of CNRS (12-18): Information, Signal Image and Vision (more than 120 laboratories and 20 industrial partners), member of the Scientific Council of the INS2I CNRS (10-14), member of steering committees of French National Research Agency. She was awarded of the Price Montpetit of the French Academy of Sciences in 2013, she is knight of the French Legion of Honour in 2015 and has a chair of the French national Artificial Intelligence Interdisciplinary Institute (3IA), 2019.
Inverse Problem in Fluorescent Microscopy and Super-Resolution
Photonic microscopy, including Widefield and confocal microscopes, is widely used in Biology platforms. As the resolution is limited by diffraction, lot of super-resolution techniques has been recently developed. They mainly relies on specific illumination or specific fluorophores but all involve numerical computations to recover a super-resolved / restored image. This course will present some of these systems and more importantly the methodology and numerical methods which are needed to compute the final image. We introduce the inverse problems in image processing with example on image deconvolution, regularization from the standard ones in image processing (e.g. TV regularization) to more complex sparse regularization involving the counting norm. Using these models, the estimation is obtained by minimizing energy criterion which can be convex or non convex, smooth or non smooth. We will then review how the optimization can be achieved and will present the main algorithms most used in image processing in the different cases. Illustrations will be given on deconvolution of confocal images, super-resolution microscopy by Single Molecule Locaization Microscopy (SMLM), Multi-Angle Total Internal Reflexion Microscopy (TIRF-MA), or recent methods using blinking molecules.