Replacement Algorithms for Fingerprint Template Update

Semi-supervised fingerprint template update is a relatively novel topic in which the problem of template representativeness with respect to unknown data has been pointed out. Current approaches to template update are based on insertion of a novel template into the gallery, and/or fusion of the template with an existing one. The automatic replacement of an existing template by a novel one is another possibility which has not yet been investigated. This paper is a first attempt to study the feasibility of using some template replacement algorithms in fingerprint verification systems. Replacement is performed online, which can be considered as a sort of "worst case" for this approach. Preliminary experiments are carried out on the four data sets of FVC2002 event.