Double-Center-Based Iris Localization and Segmentation in Cooperative Environment with Visible Illumination

Area of Science:

• Computer Science
• Biometrics
• Image Processing

Background:

• Iris recognition is a highly accurate biometric technology crucial for security applications.
• Existing iris localization methods struggle with noisy images from non-cooperative environments.
• Preprocessing tasks like iris segmentation and localization are vital for accurate iris recognition.

Purpose of the Study:

• To develop a robust end-to-end network for iris localization and segmentation.
• To overcome limitations of current methods in handling noisy and occluded iris images.
• To directly predict iris boundaries without complex post-processing.

Main Methods:

• Proposed a novel anchor-free, center-based double-circle iris localization network.
• Integrated an iris mask segmentation module for end-to-end processing.
• Introduced a concentric sampling strategy for efficient training.

Main Results:

• Achieved 84.02% box IoU and 89.15% mask IoU on the NICE-II dataset.
• Outperformed existing methods by 4.64% with 74.06% average box IoU on MICHE sub-datasets.
• Demonstrated excellent iris localization performance on four challenging datasets.

Conclusions:

• The proposed double-center-based network effectively addresses challenges in iris localization and segmentation.
• The end-to-end framework offers a significant improvement over traditional methods.
• The method shows strong potential for real-world biometric security systems.