DEALing with Image Reconstruction: Deep Attentive Least Squares

Mehrsa Pourya †, Erich Kobler ‡, Michael Unser †, Sebastian Neumayer ⋆ · 2025

Read DEALing with Image Reconstruction: Deep Attentive Least Squares by Mehrsa Pourya †, Erich Kobler ‡, Michael Unser †, Sebastian Neumayer ⋆ (2025) 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-2502-04079.txt.

Description

Mehrsa Pourya, Erich Kobler, Michael Unser, and Sebastian Neumayer rebuild images with a lean method rooted in classic maths, in work from twenty twenty-five. Top image rebuild tools now lean on huge, complex neural designs. Those designs can be hard to read, trust, or prove stable. The team asks if a classic frame can match them with less bulk. Their frame grows from Tikhonov regularization, a long-used smoothing rule.

The method refines a draft image by solving a chain of simple square problems. Learned filters pull out the key traits of each draft stage. An attention step then adjusts how hard each trait is penalised, place by place. Scores match top plug-and-play and learned rivals on the tests run. The method also stays readable, robust, and sure to settle. It bridges old maths and deep learning in one clean loop.

Catalog record

CREATORMehrsa Pourya †, Erich Kobler ‡, Michael Unser †, Sebastian Neumayer ⋆
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE46 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-04079

Browse the full library · About the Tower of Babel