Biometric Iris-Cornea Database

1.Introduction

Ocular biometrics refers to the use of features of the eye for person recognition. For instance, the unique and stable texture of the iris has been recognised as a powerful ocular biometric characteristic. We have proposed to improve biometric authentication with a multimodal ocular biometric system based on the iris pattern and the three-dimensional shape of the cornea and we have shown how the cornea can be used as a biometric trait for person recognition by using an intra-ocular fusion with iris features to improve the overall performance of the system (see reference below).

2. Description of our multimodal ocular database

Our multimodal database contains 39 subjects of different ethnic origins including 22 females and 17 males with no prior surgery or disease except common visual defects (myopia, hyperopia, astigmatism). The mean age ± SD (standard deviation) was 36.5 ± 15.6 years (range: 23-82 years). For each subject, iris images were captured with a simple set-up comprising a webcam microscope while corneal shapes were taken with a corneal topographer. The acquisitions were performed in two different sessions separated by a time interval of at least one month to evaluate within-subject variability. We took four (4) acquisitions for each cornea and ten (10) for each iris in each session for a total of 780 iris images and 312 corneal topographies.

a) Iris image acquisition


Fig. 1

Fig. 2

Fig. 3

To capture multiple images of iris, we used a simple Plugable USB 2.0 Microscope webcam (see Fig. 1). The main advantages of this camera are its very near focus, small field of view and low price. It provides up to 50X optical magnification colour images at a resolution of up to 2000 × 1600 pixels. Our basic set-up included a mount consisting of a chin rest, to position the head, and a mobile arm that can be easily manipulated to finely position the camera (see Fig. 2). The LED light source of the camera was used to illuminate the iris of the eye. In order to perform an accurate adjustment, live images were displayed on a computer screen, until the operator decided to capture acceptable images. In our setting, the sensor outputs a color image of size 640 × 480 (Fig.3) that is converted to grey level by eliminating the hue and saturation information while retaining the luminance for further processing. Some images were challenging with their dark brown color (texture is not clearly resolved) or occlusion by eyelids or eyelashes, as in a true realistic situation.

b) Cornea acquisition

The cornea is the outer transparent part of the eye, and covers nearly one fifth of the eyeball surface, with an average diameter of 11 mm and a roughly spherical shape. A non-invasive device known as corneal topographer was used for capturing the shape of each cornea. In this study, the Pentacam HR topographer was used (Fig. 4). It is a rotating Scheimpflug camera system that gives 50 images of the anterior segment of the eye in <2 s. A second camera detects all eye movement during the scan and corrects Scheimpflug images accordingly. The 50-image scan measures a total of 25,000 3D points. The Pentacam HR software then calculates a 3D model of the anterior segment and automatically provides many measurements such as elevation (height with respect to a plane perpendicular to the line of sight), pachymetry (corneal thickness) and curvature of the anterior and posterior surfaces of the cornea. In this study, the corneal shape was recorded as a uniformly spaced 141 × 141 grid of anterior surface elevations (Z), spaced by 0.1 mm intervals along the X.

Fig. 4

3.Database Organization

The database is organized as follows:

4. Download Instructions

You can download the entire database here.
For more informnations, you can contact me at nassima.kihal@polymtl.ca or kihalnassima@gmail.com

5. Publications

Please provide this reference in your publications.