Beyond MMSE: Enhancing PnP Restoration with ProxiMAP
Kenta Vert, Giacomo Meanti, Scott Pesme, Michael Arbel, Julien Mairal · 2026
Read Beyond MMSE: Enhancing PnP Restoration with ProxiMAP by Kenta Vert, Giacomo Meanti, Scott Pesme, Michael Arbel, Julien Mairal (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-16396.txt.
Description
Vert and four others show that aiming at the MAP denoiser with diffusion-model scores is problematic in Plug-and-Play restoration. Learned scores mismatch true scores, so MAP-aimed rounds converge to cartoon-like images, with better results from stopping early. ProxiMAP, a stepwise MAP estimate, keeps the current guess's leftover noise matched to the denoiser's training noise. Keeping the denoiser where its score is reliable yields early stopping that avoids the failure mode. ProxiMAP sharpens rebuilds across deblurring, inpainting, super-resolution, and phase retrieval.
Catalog record
| CREATOR | Kenta Vert, Giacomo Meanti, Scott Pesme, Michael Arbel, Julien Mairal |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 57742 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.16396