Self-Supervised Real-World Denoising by Jointly Learning Visible and Invisible Noise


Shaoyu Wang (Dalian Martime University), Changze Zhou (Dalian Maritime University), Bolin Song (Dalian Martime University), Yiyang Wang (Dalian Martime University)
The 35th British Machine Vision Conference
PDFPosterVideo (Right click to download)Supplementary

Abstract

Recently, there has been a rise in the development of self-supervised methodologies that facilitate denoising directly on real-world images without relying on clean image references. Existing self-supervised methods primarily focus on developing techniques for breaking the spatial correlation inherent in real-world noise. However, the inconsistent visibility observed in real-world noisy images, which deviates from that encountered in synthetic noisy images, has yet to be taken into account. In this paper, we propose a new perspective for self-supervised denoising on real-world noisy images, separately and jointly learning both visible and invisible noise using a single blind-spot network. To achieve this objective, a noise visibility map is estimated without relying on any ground truth or reference for the noise level, to direct the network towards focusing on the regions that exhibit similar visual performance using different strategies. Extensive experiments have been conducted to validate the superiority of our method over existing self-supervised denoisers from both quantitative and visual comparisons.

Citation

@inproceedings{Wang_2024_BMVC,
author    = {Shaoyu Wang and Changze Zhou and Bolin Song and Yiyang Wang},
title     = {Self-Supervised Real-World Denoising by Jointly Learning Visible and Invisible Noise},
booktitle = {35th British Machine Vision Conference 2024, {BMVC} 2024, Glasgow, UK, November 25-28, 2024},
publisher = {BMVA},
year      = {2024},
url       = {https://papers.bmvc2024.org/0033.pdf}
}


Copyright © 2024 The British Machine Vision Association and Society for Pattern Recognition
The British Machine Vision Conference is organised by The British Machine Vision Association and Society for Pattern Recognition. The Association is a Company limited by guarantee, No.2543446, and a non-profit-making body, registered in England and Wales as Charity No.1002307 (Registered Office: Dept. of Computer Science, Durham University, South Road, Durham, DH1 3LE, UK).

Imprint | Data Protection