DeCompress: Denoising via Neural Compression

Ali Zafari ∗ , Xi Chen ∗ , Shirin Jalali · 2025

Read DeCompress: Denoising via Neural Compression by Ali Zafari ∗ , Xi Chen ∗ , Shirin Jalali (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-2503-22015.txt.

Description

Ali Zafari, Xi Chen, and Shirin Jalali present DeCompress, a new way to clean noise from images, in this two thousand twenty-five paper. Standard learning-based denoisers need large sets of paired clean and noisy images. In fields like microscopy, clean ground truth images often cannot be had at all. Some methods learn from noisy data alone, but they still need many samples and heavy training.

DeCompress draws on neural compression instead. It needs no clean images and no large training set. A single noisy image is enough for it to work. The authors say it also resists overfitting. Their results point to denoising that fits small labs and rare imaging tasks. The paper speaks to imaging researchers who lack the data that big models demand.

Catalog record

CREATORAli Zafari ∗ , Xi Chen ∗ , Shirin Jalali
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE24 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-22015

Browse the full library · About the Tower of Babel