Scanned documents forensics: detecting inserted characters through noise and chromatic artifacts

Marina Gardella, Julieta Umpierrez, Antoine Tadros, Seginus Mowlavi, Natalia Bottaioli, Diego Belzarena, Gabriele Facciolo, Roy He, Jean-Michel Morel, Rafael Grompone Von Gioi; Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops, 2025, pp. 7574-7584

Abstract

Document forgery detection plays a crucial role in safeguarding the integrity of various sectors, including accounting, insurance, finance, law enforcement, and national security. The ability to distinguish between authentic and counterfeit documents is key to ensure trust in transactions, maintain regulatory compliance, and prevent fraudulent activities. Despite its importance, the field of document forgery detection remains underdeveloped, especially compared to the advances made in image forgery detection. In this work, we present the first benchmarking results of state-of-the-art image forgery detection methods applied to forged documents, and we demonstrate that these methods underperform in this context. To address this gap, we introduce two novel approaches specifically designed for document forgery detection. These approaches analyze anomalies in noise distribution and chromatic artifacts in scanned documents to identify inserted characters. The source code is available in https://github.com/julietaumpierrez/Scanned-documents-forensics

Related Material

[pdf]
[bibtex]
@InProceedings{Gardella_2025_ICCV, author = {Gardella, Marina and Umpierrez, Julieta and Tadros, Antoine and Mowlavi, Seginus and Bottaioli, Natalia and Belzarena, Diego and Facciolo, Gabriele and He, Roy and Morel, Jean-Michel and Von Gioi, Rafael Grompone}, title = {Scanned documents forensics: detecting inserted characters through noise and chromatic artifacts}, booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops}, month = {October}, year = {2025}, pages = {7574-7584} }