A New approach for fingerprint recognition using Earth Mover’s Distance

Main Article Content

B.K. Tripathy
P.V.S.S.R.Chandramouli, Bimalendu Ranajit

Abstract

Fingerprint recognition has been widely used in information security, personal identification etc. for its uniqueness and invariance. Efficient features extraction from input image is one of the important components of the fingerprint image recognition. Till now various algorithm have been proposed by many researchers. Out of the two widely followed approaches for fingerprint recognition, the minutia-based technique represents the fingerprints by its local features like terminations and bifurcations. This approach has been intensively studied and is also the backbone of the current available fingerprint recognition products. In this paper, we developed a fingerprint recognition concept, which uses the regular texture regions of fingerprint image that can be successfully represented by co-occurrence matrices. We first extract the features based on certain characteristics of the co-occurrence matrix. Simultaneously, the histogram of each image with certain bins and texture features is also extracted. These histogram values obtained for both the existing finger prints and the fingerprint to be tested are fed to the EMD (Earth mover’s distance technique) algorithm to measure the distances. These measures with a certain threshold value are used for the matching. Experimental results show that our approach is very efficient in recognizing both poor quality images and variants of the same image.

 

Key words: Fingerprints, Gabor filter, EMD, gray level co-occurrence matrix (GLCM)

Downloads

Download data is not yet available.

Article Details

Section
Articles