A scalable estimator of space varying blurs Application in super-resolution

Valentin Debarnot (1) , Thomas Mangeat (2) , Pierre Weiss (3)

1 IMT - Institut de Mathématiques de Toulouse UMR5219
2 LBCMCP - Laboratoire de Biologie Cellulaire et Moléculaire du Contrôle de la Prolifération
3 CNRS - Centre National de la Recherche Scientifique

We propose a scalable method to find a low dimensional subspace of spatially varying blur operators, given a set of noisy observations. This aims to improve the identifiability in blind inverse problems. We then show how to use this subspace estimator to train deep learning architectures for solving super-resolution problems in single molecule localization microscope.

Domaines: Traitement du signal et de l'image [eess.SP]

Valentin Debarnot, Thomas Mangeat, Pierre Weiss. A scalable estimator of space varying blurs Application in super-resolution. SPARS, Jul 2019, TOULOUSE, France. ⟨hal-02374520⟩