Orateur : Jérôme Idier
Affiliation : Equipe ADTSI, IRCCyN, Nantes
Titre : Etude théorique de la capacité de super-résolution en microscopie de fluorescence à éclairement de type speckle
Résumé : A préciser
Orateur : Marco Bevilacqua
Affiliation : Labri/IMB
Title: Example-based algorithms for single-image super-resolution
Abstract : Super-resolution (SR) refers to a family of methods that aim at increasing the resolution of given images to a greater extent than traditional image processing techniques. In particular, example-based SR methods typically employ tools from the machine learning field, to estimate patch-by-patch the high-resolution (HR) version of a single low-resolution (LR) input image. In order to perform this task, they make use of a dictionary of "examples" (pairs of LR/HR patches), which can either be trained from external images (leading to an external dictionary), or derived by looking for self-similarities within the input image itself (we then have an "internal" dictionary).
In this talk, after providing some context, I will describe two distinct example-based SR algorithms, one for each kind of dictionary, developed as part of my PhD thesis work.