Marcelo Pereyra | Heriot-Watt University

Marcelo Pereyra | BOLT

Maxwell Insitute for Mathematical Sciences
School of Mathematical and Computer Sciences, Heriot-Watt University

BOLT: Bayesian model selection and calibration for computational imaging

Summary

BOLT is a 3-year project funded by UKRI EPSRC (EP/T007346/1) to develop new mathematical methods and computer algorithms for solving inverse problems related to computational imaging, with a particular focus on problems that are blind, semi-blind, or unsupervised.

The aim is to provide computationally efficient and theoretically rigourous tools for automatic model selection and calibration, directly from the observed raw data, without using ground truth data, and with minimum expert supervision. Such tools would significantly simplify deploying computational imaging methodology and increase robustness to variability in imaging setups and scenes.

During the project, we will investigate two applications to remote sensing: hyperspectral image pansharpening and radio-interferometric astronomical image reconstruction.

PhDs and Postdocs

Dr David Thong (postdoc)
Charlesquin Kemajou (phd student)

Collaborators

Prof. Jean-François Giovannelli, University of Bordeaux (web)
Prof. Jason McEwem, MSSL UCL (web)
Dr. Xiaohao Cai, Southampton (web)
Prof. Jean-Yves Tourneret, University of Toulouse (web)
Prof. Steve McLaughlin, Heriot-Watt University (web)
Dr. Alain Durmus, Centre Borelli, ENS Paris-Saclay (web)
Prof. Eric Moulines, Ecole Polytechnique (web)