Palmprint identification using hashing of 2D-Log Gabor filter, Phase Congruency and Haralick features

In palmprint identification system, a particular person´s biometric template is compared against all images in the database to identify the person. The matching procedure in large database degrades the performance of biometric system. The proposed method consists of three stages feature extraction, hashing generation and querying. 2D-Log Gabor, Phase Congruency and Haralick feature methods are used for feature vector construction. In hash generation, all the templates in the database are indexed according to the orientation features. The hash table, size table and offset table are constructed to speed up the matching process. In querying stage, test image features are extracted and matched using the hash table. The experiments are evaluated on a standard database (PolyU Database), and the results show that significant improvement in terms of recognition accuracy and error rates.