Efficient features for smartphone-based iris recognition

Ritesh Vyas,Tirupathiraju Kanumuri,Pawan Dubey,Gyanendra Sheoran

doi:10.3906/elk-1809-98

Ritesh Vyas, Tirupathiraju Kanumuri + Show 2 more

Open Access

https://doi.org/10.3906/elk-1809-98

Copy DOI

Abstract

Iris recognition has widely been used in personal authentication problems. Recent advances in iris recognition through visible wavelength images have paved the way for the use of this technology in smartphones. Smartphone-based iris recognition can be of significant use in financial transactions and secure storage of sensitive information. This paper presents a hybrid representation scheme for iris recognition in mobile devices. The scheme is called hybrid because it firstly makes use of Gabor wavelets to reveal the texture present in the normalized iris images, and then extracts statistical features from different partitions of Gabor-processed images. The standard mobile-iris database, called MICHE, is used for investigating the performance of the proposed approach. The comparison of the proposed approach with other widespread iris recognition approaches proves its efficacy.

Highlights

Iris has been the most popular biometric trait because of its uniqueness, permanence, and noninvasiveness [1]
Phase I: Evaluation of distance metrics This phase is aimed at selecting appropriate distance function for the addressed problem of mobile iris biometrics
This paper presents a hybrid feature extraction approach to be used in mobile iris biometrics

Summary

Introduction

Iris has been the most popular biometric trait because of its uniqueness, permanence, and noninvasiveness [1]. The present work reports a hybrid feature extraction technique for mobile iris biometrics. Each Gabor-processed image is partitioned into subblocks to take the local features of iris into account. For second level of partitioning, the dimension of subblocks are chosen to have increased horizontal and vertical resolutions This increased resolution of subblocks aid in countering the effects of eyelid and eyelash occlusions, present in the normalized iris image. The eye region is detected using the Viola–Jones framework [30, 31] This framework is based on extraction of rectangular features from the integral images of input eye image.

Proposed approach

5: Process I1 with 2D Gabor wavelet through convolution operation

Phase I

Phase II

Conclusions