Skip to content. | Skip to navigation

Personal tools

Sections

UMR 5672

logo de l'ENS de Lyon
logo du CNRS
You are here: Home / Seminars / Machine Learning and Signal Processing / A linear MMSE approach to blind inverse problems in imaging

A linear MMSE approach to blind inverse problems in imaging

Nathan Buskulic (Postdoctoral researcher at the Machine Learning center of Genova (MaLGa))
When Nov 24, 2026
from 01:00 to 02:00
Where M7 101
Add event to calendar vCal
iCal

Abstract: Blind inverse problems arise frequently in scientific imaging such as computational microscopy or astronomy, where perturbations like atmospheric turbulence create uncertainty about the observation process and the forward operator. Solving blind inverse problems comes with additional challenges due to the existence of intrinsic ambiguities (e.g., scale or translation), which makes traditional non-blind reconstruction methods unfit for the task. I will show how a linear MMSE estimator, akin to a Wiener/Kolmogorov filter, can be adapted for blind reconstruction and how it can be equipped with various theoretical guarantees. I will also discuss how it can be used in practice to make Maximum A Posteriori approaches more robust to ambiguous solutions.