Palmprint feature extraction approach using nonsubsampled contourlet transform and orthogonal moments

Abstract

In recent usage of internet in financial and information transaction, authentication becomes necessity for authorized access. Palmprint recognition is a widely accepted biometric authentication among various biometrics methodologies. The enormous feature content and less acquisition cost make it more reliable and user friendly. Texture is one of the vital features in biometric recognition applications. Though many statistical methods are available to extract the texture, non-subsampled contourlet transform is employed in this work as a primary step to extract the directional frequency information content in the palmprint, followed by the statistical moment extraction. In addition to using the Zernike moments as texture descriptors, they are effectively used in reducing the dimensionality of contourlet coefficients. Since Zernike moments are inherently orthogonal and rotation invariant, they are more suitable for palmprint recognition.