XFinger-Net: Pixel-Wise Segmentation Method for Partially Defective Fingerprint Based on Attention Gates and U-Net

Area of Science:

• Computer Science
• Biometrics
• Image Processing

Background:

• Partially defective fingerprint images (PDFIs) present significant challenges to the accuracy and performance of automated fingerprint identification systems (AFIS).
• Existing segmentation methods often rely on features like ridge orientation and frequency, which can be insufficient for complex or degraded fingerprint data.
• Accurate segmentation is crucial for enhancing image quality and improving the reliability of AFIS.

Purpose of the Study:

• To introduce XFinger-Net, a novel deep learning-based method for segmenting partially defective fingerprint images.
• To improve the quality and performance of PDFIs within AFIS.
• To address the limitations of current segmentation techniques in handling degraded fingerprint data.

Main Methods:

• Developed XFinger-Net, a U-Net architecture incorporating an attention gate to suppress irrelevant regions and focus on pertinent fingerprint features.
• Implemented pixel-level segmentation to precisely delineate fingerprint areas.
• Utilized non-blocking fingerprint images as input to preserve global characteristics and contextual information.

Main Results:

• The proposed XFinger-Net achieved a very good segmentation effect on a self-made test dataset of fingerprint images.
• The attention gate mechanism effectively suppressed uncorrelated regions, leading to more focused segmentation.
• Pixel-level segmentation preserved fine details crucial for identification.

Conclusions:

• XFinger-Net demonstrates significant potential for enhancing the segmentation of partially defective fingerprint images.
• The method offers a promising solution for improving the performance and accuracy of automated fingerprint identification systems.
• Further validation on diverse and challenging fingerprint datasets is warranted.