Rotation Invariant Feature Extraction of Handwritten Signature Images

This paper presents extraction of rotation invariant features of handwritten signature images for classification. Initially, curvelet transform was applied to all signature image in the database. The mean and standard deviation values of each curvelet sub bands were used as features. Rotation invariance was obtained by applying cycle shift around the total spectral energy values of curvelet sub bands. Then, the classification process was carried out with the obtained features. The performance was compared with the method without cycle shift. As a result, it has been determined that the presented method gives the most successful accuracy using the support vector machine classifier.