Forgery Localization Via Extracting Generic Features And Multiple Prior Fusion
Journal of Intelligent Computing and Networking · 31.05.2026 · 2 (2) · 1-13
DOI: 10.64509/jicn.22.94 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Intelligent Computing and Networking, online erschienen am 31.05.2026. Verfasst von Jiaqi Zhang und 4 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref). Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Nowadays, it is easy to create fake images using image editing software, which poses many security risks. Many current deep learning approaches tend to mix multiple types of faked samples for training to improve the model's generality. This ignores the unique nature of the different forged image types. In this paper, we provide a new perspective on training strategies for the generality of forensic tasks. Our analysis and experiments suggest that, under the conditions of this study, copy-move samples may be more effective in learning generic features than splicing samples. To explore this generic feature, we trained using only copy-move samples and tested both splicing and copy-move samples. In addition, we observed that original images of different quality underwent the same forgery process and did not produce precisely the same forgery cues, even though they had the same semantic information. Therefore, in order to improve the generalisation of the method to different datasets, three copy-move datasets were created based on the different qualities of the original images. Instead of mixing the three copy-move datasets directly for training, we use a multi-prior fusion strategy for training. The effectiveness of our proposed method was demonstrated through experimental testing on the public datasets.
Zitieren
APA
Zhang, J., Jian, L., Ma, Z., Ma, J., & Huang, X. (2026). Forgery Localization Via Extracting Generic Features And Multiple Prior Fusion. Journal of Intelligent Computing and Networking, 2(2), 1-13. https://doi.org/10.64509/jicn.22.94
Vancouver
Zhang J, Jian L, Ma Z, Ma J, Huang X. Forgery Localization Via Extracting Generic Features And Multiple Prior Fusion. Journal of Intelligent Computing and Networking. 2026;2(2):1-13. doi:10.64509/jicn.22.94
BibTeX
@article{Zhang2026,
author = {Zhang, Jiaqi and Jian, Liqiong and Ma, Ziping and Ma, Jinlin and Huang, Xiaoshuai},
title = {Forgery Localization Via Extracting Generic Features And Multiple Prior Fusion},
journal = {Journal of Intelligent Computing and Networking},
year = {2026},
volume = {2},
number = {2},
pages = {1--13},
doi = {10.64509/jicn.22.94}
}
EUVIMED