Hand-written signatures represent nowadays the primary mechanism for authorization in legal transactions. Accordingly, the search for algorithms that allow an effective automatic comparison between images of hand-written signatures turns out to be of the greatest importance. Studies in this area have to face some problems due to intrapersonal variations and interpersonal differences, that make it necessary to analyse signatures as complete images and not as letters and words put together. Therefore, only an extensive analysis of these geometric objects can suggest the most suitable features that have to be evaluated. Such features can be extracted at local or global level of the image and then have to be compared with an appropriate classifier. Moreover, the geometric features often need to be processed on remote clusters, still guaranteeing their privacy. So, we identified some features that could be protected using a suitable form of homomorphic encryption
Geometric features for hand-written signatures / Pellegrini, Chiara; Rimoldi, Anna; Sala, Massimiliano. - STAMPA. - 84:(2014), pp. 117-134. (Intervento presentato al convegno SAGA 2012 tenutosi a Trento nel 2012) [10.1007/978-1-4471-6461-6_8].
Geometric features for hand-written signatures
Pellegrini, Chiara;Rimoldi, Anna;Sala, Massimiliano
2014-01-01
Abstract
Hand-written signatures represent nowadays the primary mechanism for authorization in legal transactions. Accordingly, the search for algorithms that allow an effective automatic comparison between images of hand-written signatures turns out to be of the greatest importance. Studies in this area have to face some problems due to intrapersonal variations and interpersonal differences, that make it necessary to analyse signatures as complete images and not as letters and words put together. Therefore, only an extensive analysis of these geometric objects can suggest the most suitable features that have to be evaluated. Such features can be extracted at local or global level of the image and then have to be compared with an appropriate classifier. Moreover, the geometric features often need to be processed on remote clusters, still guaranteeing their privacy. So, we identified some features that could be protected using a suitable form of homomorphic encryptionI documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione