HAL Université des Antilles

Hyper-spectral Analysis for Automatic Signature Extraction

Muhammad Imran Malik ¹ Sheraz Ahmed ¹ Faisal Shafait ² Ajmal Saeed Mian ² Christian Nansen ² Andreas Dengel ¹ Marcus Liwicki ^{1, 3}

1 DFKI - Deutsches Forschungszentrum für Künstliche Intelligenz GmbH

2 UWA - The University of Western Australia

3 Department of Informatics

Abstract : Signatures are one of the most accessible and prevailed ways of authenticating documents. Over the last many years, a large number of signature verication systems have been reported. A common assumption in nearly all of these system is that the signatures are available readily extracted from documents. In this paper we provide a detailed literature survey on the subject and argue that pre-extracted signatures are not always available especially in forensic cases. Furthermore, we present a novel system of automatically extracting signatures from documents with the help of hyper-spectral imaging. Initial experiments reveal that the proposed idea possess great potential to form a baseline signature extraction system above whom any signature verication system can be adjuncted for signature verication.

Document type :

Conference papers

Domain :

Computer Science [cs]

Complete list of metadatas

Cited literature [31 references] Display Hide Download

https://hal.univ-antilles.fr/hal-01165888
Contributor : Philippe Simon <>
Submitted on : Monday, June 22, 2015 - 8:47:28 PM
Last modification on : Thursday, May 2, 2019 - 2:16:11 PM
Long-term archiving on: Tuesday, September 15, 2015 - 7:57:30 PM

IGS_2015_submission_18.pdf

Files produced by the author(s)