Palmprint Recognition Using Radon Transform

Abstract: In this paper we have proposed a novel person identification system that uses palmprint features extracted by using Radon transform. The palmprint image has been viewed as a texture image. The local features from the extracted ROI of a palmprint represent the texture information present in the palmprint image in better sense. Radon transform computes the line integral along parallel paths in a certain direction. Here we have computed line integrals of the palmprint image at 60 different directions ranging from 0-180 degree with the interval of 3 degree. These feature vectors of size 185×60 have been used as the feature map. Performance of the proposed algorithm has been tested using PolyU database and the in-house generated database. Maximum recognition accuracy of 95.60% for PolyU palmprint database and that of 97.20% for the in-house generated palmprint images have been achieved with the algorithm. The algorithm is computationally efficient as the testing time required is 0.86 sec.

Keywords: palmprints, Radon transform, region of interest, Euclidian distance.