Improved feature based offline signature enhancement and verification

Several recent advancement in field of technology has created a various range of systems or applications that requires reliable and efficient recognition schemes which can determine the identity of the users and only authorized and legitimate persons are allowed to use the service or applications. Signature Verification is the most widely used biometrics for authentication purpose. It is a method which is legally accepted and is used in many areas like banking, e-business, access control etc. Guided filter which is a very new edge preserving smoothing algorithm has been used for quality enhancement of signature images. This paper proposes signature verification algorithm where 6 different features of the signature are used to determine its legitimacy. The signature database used for this work consist of above 3000 signatures from different writers and consists of both genuine and forged signatures. The signatures were first captured offline and then converted into digital form. The algorithm can detect both Random and skilled forgeries. The proposed algorithm achieves an Average Error Rate(AER) of 1.8% when only skilled forgeries of a writer is considered.