Signature verification using Directional and Textural features

Biometric identification technique like offline signature verification and recognition is now a day considered as one of the important personal identification method used to identify the individual. Feature extraction is the best technique which preserves the essential information of the input image. In this paper we propose offline signature verification based on Transform domain feature such as gradient, coherence and dominant local orientation. The acquired image is resized to bring all the signatures into a uniform size. The images are thinned using morphological process. The DWT technique is applied on signature images to get LL, LH, HL and HH subbands. The directional information feature is computed from the subbands. The directional features and textural features are concatenated to form the feature vector. The Feed Forward ANN tool in MATLAB is used for classification and verification. The results of False Rejection Rate (FAR), False Acceptance Rate (FAR) and Total Success Rate (TSR) are obtained for GPDS-960 database. A total of 360 images are used for training and testing. It is observed that the values of FRR, FAR and TSR are improved compared to the existing algorithms.