Smartphone-based Iris Recognition through High-Quality Visible Spectrum Iris Capture

TL;DR

This work tackles the challenge of smartphone-based iris recognition in the visible spectrum by delivering an end-to-end Android solution that automatically captures high-quality VIS iris images and performs on-device segmentation. It combines a compact YOLOv3-Tiny eye/iris detector with a lightweight Ghost-Attention U-Net (G-ATTU-Net) segmentation model, all while enforcing ISO/IEC 29794-6 image quality standards. The authors introduce the CUVIRIS dataset (VIS and NIR images from 47 subjects) and demonstrate strong verification performance: VIS TAR of 96.57% (left) and 96.81% (right), NIR TAR of 97.95%, and cross-spectral TAR around 96.2%, across varying distances and iris colors. The results support the feasibility of reliable cross-spectral iris recognition on mobile devices and provide a public dataset and a lightweight segmentation model suitable for real-time deployment, advancing smartphone security and biometrics research.

Abstract

Iris recognition is widely acknowledged for its exceptional accuracy in biometric authentication, traditionally relying on near-infrared (NIR) imaging. Recently, visible spectrum (VIS) imaging via accessible smartphone cameras has been explored for biometric capture. However, a thorough study of iris recognition using smartphone-captured 'High-Quality' VIS images and cross-spectral matching with previously enrolled NIR images has not been conducted. The primary challenge lies in capturing high-quality biometrics, a known limitation of smartphone cameras. This study introduces a novel Android application designed to consistently capture high-quality VIS iris images through automated focus and zoom adjustments. The application integrates a YOLOv3-tiny model for precise eye and iris detection and a lightweight Ghost-Attention U-Net (G-ATTU-Net) for segmentation, while adhering to ISO/IEC 29794-6 standards for image quality. The approach was validated using smartphone-captured VIS and NIR iris images from 47 subjects, achieving a True Acceptance Rate (TAR) of 96.57% for VIS images and 97.95% for NIR images, with consistent performance across various capture distances and iris colors. This robust solution is expected to significantly advance the field of iris biometrics, with important implications for enhancing smartphone security.

Figures (12)

• Figure 1: Automatic Focal Point Adjustment. The left column shows out-of-focus eye images before the focal point adjustment, while the right column shows the corresponding images after the focus is automatically adjusted to the iris region.
• Figure 2: The architecture of the proposed G-ATTU-Net model. Each dark blue block represents a multi-channel feature map produced by a Ghost Module, with the numbers above indicating the number of channels and the vertical numbers on the left indicating the spatial dimensions (height x width). The red blocks indicate Attention Blocks used to refine features before concatenation in the decoder path.
• Figure 3: The input images and segmentation results using both the standard Attention U-Net and the GAtt-UNet. The first column shows the red channel images extracted from RGB inputs after cropping the eye region using the YOLOv3-Tiny eye and iris detection model from the CUVIRIS dataset. The second column contains the ground truth masks for the iris region. The third column displays the segmentation results from the standard ATT-UNet, and the fourth column shows the results from the G-ATTU-Net.
• Figure 4: (a) User interface for inputting participant details and (b) camera preview displaying detected eye and iris bounding boxes with real-time user feedback, ensuring accurate and focused image capture.
• Figure 5: Process flow of the Android application for capturing and processing iris images.