Palmprint based verification system robust to rotation, scale and occlusion

This paper proposes an efficient palmprint based verification system which is robust to rotation, scale and occlusion. Images are obtained using a flat bed scanner. Scale invariant feature transform (SIFT) operator is used to extract features from the palmprint. Nearest neighbor ratio method is used to determine the similarity between extracted features of live and enrolled palmprints and to make matching decision. The proposed system has been tested using three databases-IITK database having 549 hand images, CASIA database with 5239 hand images and PolyU database of size 7752. Accuracy of the proposed system is found to be 99.97% with FAR of 0.06% in case of IITK database, while for CASIA and PolyU database is more than 99%. Further the robustness of the system with respect to scale rotation and occlusion has been studied.