Invariant Feature Extraction from Fingerprint Biometric Using Pseudo Zernike Moments
C. Lakshmi Deepika, Arumugam Kandaswamy, C. Vimal, B. S. Sathish · 2010
To represent the large amount of data in the biometric images an efficient feature extraction method is needed. Further biometric image acquisition is subject to deforming processes such as rotation, translation and scaling. Hence it is also required that the image representation be invariant to the deformations and sustain the discriminating features. Considering the trade off between the discriminating power and the invariants, moments are a very qualifying object descriptor. In this paper, we have used Pseudo Zernike moments to create invariant feature vectors for the Finger print biometric. We have used the Bayesian classifier to validate our usage of moments. The accuracy of the system was found to be 96.89% on using lower order moments.