Article Dans Une Revue IEEE Signal Processing Letters Année : 2025

PlaSo: Unsupervised Residual Plug and Play for Image Super-Resolution

Résumé

This letter addresses the problem of single image super resolution using Plug-and-play (PnP) frameworks incorporating with residual learning. We propose a novel PnP algorithm, PlaSo, which integrates residual learning mechanisms into diffusion processes to mitigate common challenges such as hallucination artifacts and over-smoothing. Furthermore, by initializing the reverse sampling process from the degraded input rather than random noise, PlaSo achieves faster convergence. Experiments demonstrate that PlaSo consistently outperforms state-of-the-art methods, both qualitatively and quantitatively.

Fichier principal
Vignette du fichier
Plaso.pdf (5.03 Mo) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)
Licence

Dates et versions

hal-05194052 , version 1 (31-07-2025)
hal-05194052 , version 2 (08-10-2025)

Licence

Identifiants

Citer

Xuan-Hieu Le, Denis Kouamé, Duong-Hung Pham. PlaSo: Unsupervised Residual Plug and Play for Image Super-Resolution. IEEE Signal Processing Letters, 2025, 32, pp. 4059-4063. ⟨10.1109/LSP.2025.3620784⟩. ⟨hal-05194052v1⟩
3170 Consultations
607 Téléchargements

Altmetric

Partager

  • More