Hand Geometry- and Palmprint-Based Biometric System with Image Deblurring

Abstract

Due to peg-free system for image acquisition, it is obvious to have a motion of user’s hand. Motion activity creates the blurring effect in captured images, which leads to the false identification of a person. In such cases, image deblurring is required to restore the image. This paper proposes hand geometry- and palm texture-based person identification system with motion blur deblurring of palm images. The deblurring process may generate the ringing artifacts. The quality of the restoration of a motion blurred image is depending on the estimation of motion blur parameters, blur angle and blur length. In this paper, Radon transform and cepstral method have been used for the blur parameter estimation. Point spread function (PSF) can be constructed using blur parameters. For the deblurring of the palm images, L2/TV regularization based on augmented Lagrangian method has been used. It deblurs the image with the removal of ringing artifacts. Restored images are further used in the biometric system to analyze the robustness of the system against the motion blur. This system is tested on a database collected from 100 people. 99.5% genuine acceptance rate (GAR) is achieved for bimodal system against motion blur for discrete cosine transform (DCT)-based features. Equal error rate (EER) achieved is 1.1719.