BUFF: Bayesian Uncertainty Guided Diffusion Probabilistic Model for Single Image Super-Resolution

Zihao He, Shengchuan Zhang, Runze Hu, Yunhang Shen, Yan Zhang · 2025

Read BUFF: Bayesian Uncertainty Guided Diffusion Probabilistic Model for Single Image Super-Resolution by Zihao He, Shengchuan Zhang, Runze Hu, Yunhang Shen, Yan Zhang (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-2504-03490.txt.

Description

Zihao He and co-authors present a method to sharpen a photo by using doubt maps to guide the cleanup, in a paper from 2025. Diffusion models can rebuild fine detail, but a plain noise model often fails on complex textures. That leads to blur and odd marks in hard areas.

Their model first builds a map of where it is unsure, using a Bayesian network. That map then sets how much noise to add in each place, so detail rich zones get gentler treatment. Tests on a standard data set beat base methods on two quality scores. Visual results showed strong gains in areas with fine texture and edges, and good stability under complex noise.

Catalog record

CREATORZihao He, Shengchuan Zhang, Runze Hu, Yunhang Shen, Yan Zhang
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE34 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-03490

Browse the full library · About the Tower of Babel