On-line signature verification based on correlation image

In this paper we proposed a novel method for signature verification. Different from the conventional approaches, we suppose the sample points of the signature have some kind of correlation with each other, and the correlations contain the habitual features of the signature, which can be used to confirm the identity of the signer. In the proposed method, we first generate the matrix of the correlations of the sample points of the signature and convert the matrix into an intensity image, then use image matching approach to verify signature. The proposed method is validated on the SVC 2004 database and inspiring results are obtained.