Maximum likelihood estimation of regularisation parameters in high-dimensional inverse problems: an empirical Bayesian approach
When |
Jan 28, 2021
from 02:30 to 03:30 |
---|---|
Contact Name | Pierre Borgnat |
Attendees |
Valentin De Bortoli |
Add event to calendar |
vCal iCal |
Title: Maximum likelihood estimation of regularisation parameters in high-dimensional inverse problems: an empirical Bayesian approach.
Abstract: Many imaging problems require solving an inverse problem that is ill-conditioned or ill-posed. Imaging methods typically address this difficulty by regularising the estimation problem to make it well-posed. This often requires setting the value of the so-called regularisation parameters that control the amount of regularisation enforced. These parameters are notoriously difficult to set a priori, and can have a dramatic impact on the recovered estimates. In this work, we propose a general empirical Bayesian method for setting regularisation parameters in imaging problems that are convex w.r.t. the unknown image. The proposed algorithm uses the same basic operators as proximal optimisation algorithms, namely gradient and proximal operators, and it is therefore straightforward to apply to problems that are currently solved by using proximal optimisation techniques. Our methodology is demonstrated on a range of experiments including image denoising, non-blind image deconvolution, and hyperspectral unmixing, using synthesis and analysis priors involving the L1, total-variation, total-variation and L1, and total-generalised-variation pseudo-norms.
The presentation is based on articles published at SIAM Imaging:
Room : Webconf3