Multi-Scale Wavelet-Enhanced U-Mamba Network for Image Forgery Localization
Information · 26.05.2026 · 17 (6) · 526
DOI: 10.3390/info17060526 ↗Kurz gefasst
Fachartikel in der Zeitschrift Information, online erschienen am 26.05.2026. Verfasst von Bing Qi und 2 weiteren Autor:innen. Die Arbeit nennt 59 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
The widespread availability of image editing tools and generative AI has made image forgery more accessible and deceptive, demanding more advanced localization techniques. Existing CNN-based methods are limited by local receptive fields, struggling with long-range dependencies, while Transformers suffer from the quadratic complexity of self-attention, hindering practical deployment. Moreover, effectively utilizing multi-scale features remains challenging. To address these challenges, we propose a Multi-scale Wavelet-enhanced U-Mamba network (MWEU-Mamba). The proposed framework employs a Mamba-based state space model as the backbone to achieve global contextual modeling with linear complexity. A wavelet enhancement module is introduced to integrate spatial–frequency representations, improving sensitivity to subtle manipulation traces across scales, while a channel attention mechanism further amplifies forgery-relevant feature responses. Extensive experiments on six public benchmark datasets (e.g., CASIA and Coverage) demonstrate that the proposed method achieves state-of-the-art performance on multiple datasets in terms of pixel-level F1-score.
Zitieren
APA
Qi, B., Ye, C., & Ding, Y. (2026). Multi-Scale Wavelet-Enhanced U-Mamba Network for Image Forgery Localization. Information, 17(6), 526. https://doi.org/10.3390/info17060526
Vancouver
Qi B, Ye C, Ding Y. Multi-Scale Wavelet-Enhanced U-Mamba Network for Image Forgery Localization. Information. 2026;17(6):526. doi:10.3390/info17060526
BibTeX
@article{Qi2026,
author = {Qi, Bing and Ye, Chunyang and Ding, Yuliang},
title = {Multi-Scale Wavelet-Enhanced U-Mamba Network for Image Forgery Localization},
journal = {Information},
year = {2026},
volume = {17},
number = {6},
pages = {526},
doi = {10.3390/info17060526}
}
EUVIMED