Multispectral scleral patterns for ocular biometric recognition

Abstract

Biometrics is the science of recognizing people based on their physical or behavioral traits such as face, fingerprints, iris, and voice. Among the various traits studied in the literature, ocular biometrics has gained popularity due to the significant progress made in iris recognition. However, iris recognition is unfavorably influenced by the non-frontal gaze direction of the eye with respect to the acquisition device. In such scenarios, additional parts of the eye, such as the sclera (the white of the eye) may be of significance. In this dissertation, we investigate the use of the sclera texture and the vasculature patterns evident in the sclera as potential biometric cues. Iris patterns are better discerned in the near infrared spectrum (NIR) while vasculature patterns are better discerned in the visible spectrum (RGB). Therefore, multispectral images of the eye, consisting of both NIR and RGB channels, were used in this work in order to ensure that both the iris and the vasculature patterns are successfully imaged. The contributions of this work include the following. Firstly, a multispectral ocular database was assembled by collecting high-resolution color infrared images of the left and right eyes of 103 subjects using the DuncanTech MS 3100 multispectral camera. Secondly, a novel segmentation algorithm was designed to localize the spacial extent of the iris, sclera and pupil in the ocular images. The proposed segmentation algorithm is a combination of regionbased and edge-based schemes that exploits the multispectral information. Thirdly, different feature extraction and matching method were used to determine the potential of utilizing the sclera and the accompanying vasculature pattern as biometric cues. The three specific matching methods considered in this work were keypoint-based matching, direct correlation matching, and minutiae matching based on blood vessel bifurcations. Fourthly, the potential of designing a bimodal ocular system that combines the sclera biometric with the iris biometric was explored. Experiments convey the efficacy of the proposed segmentation algorithm in localizing the sclera and the iris. The use of keypoint-based matching was observed to result in the best recognition performance for the scleral patterns. Finally, the possibility of utilizing the scleral patterns in conjunction with the iris for recognizing ocular images exhibiting non-frontal gaze directions was established. ACKNOWLEDGMENTS I wish to express my gratitude to my advisor Dr. Arun Ross for his guidance, knowledge and invaluable assistance throughout my research work. He has made available his support whenever I needed it. I owe my deepest gratitude to my supervisor, Dr. Lawrence Hornak for his support during the duration of my studies. Special thank you to the members of the supervisory committee Dr. Donald Adjeroh, Dr. Xin Li and Dr. Odom Vernon that contributes to the success of this study. I would also like to thank my husband Musat, and my children, Irina and Tudor, for their love and support.

A Comprehensive Approach for Multi Biometric Recognition Using Sclera Vein and Finger Vein

“…For initial log in uses sclera blood veins as it is unique to each individual and it is more accurate than any other biometric. The experimental upshots show that sclera recognition is a promising new biometrics for positive human ID [18]. A new method for sclera segmentation which works for both color and grayscale images is proposed and, designed a Gabor wavelet-based sclera pattern enhancement method to emphasize and binarize the sclera vessel patterns.…”

Interactive Growcut Multi-Label N-D Segmentation Based Algorithm for SCLERA Segmentation

“…Unfortunately, these methods results in a multitude of disconnected segments from which it is difficult to select the proper boundaries of the sclera region. The region growing and split and merge method were also employed without satisfactory results due to the large variation and intensity values across the sclera region [16]. Further, due to improper illumination, high frequency or blend images, overlapping of region of interest, segmentation based on sclera region are less robust and do not provide good results.…”

“…The researcher was not being able to test the method by itself but by presenting the previous study [16] conducted. The previous study shows that employing K-means algorithm in sclera segmentation, shows even promising performance but it was found out that it erroneously labels portion of the sclera as being the iris while on the other hand applying interactive GrowCut shows better performance by providing the result presented by previous research and the current result of the experiment as shown in the figure below.…”

“…The previous study shows that employing K-means algorithm in sclera segmentation, shows even promising performance but it was found out that it erroneously labels portion of the sclera as being the iris while on the other hand applying interactive GrowCut shows better performance by providing the result presented by previous research and the current result of the experiment as shown in the figure below. a) Sample segmentation using K-means [16] b) Sample segmentation using Interactive Growcut algorithm [29] Fig. 6.…”

In vivo oximetry of human bulbar conjunctival and episcleral microvasculature using snapshot multispectral imaging

“…Repeated semi-automatic tracking demonstrated negligible 314 variation in ODR (a standard deviation of <0.5% in 10 repeated measurements). 315 and episcleral vasculature (Crihalmeanu and Ross, 2012). 318 319…”