Study of Full-View Finger Vein Biometrics on Redundancy-compensated 3D Reconstruction and Frequency-Spatial Coupling...

Area of Science:

• Biometrics
• Computer Vision
• Pattern Recognition

Background:

• Full-view 3D finger vein (3DFV) biometrics offers advanced security but faces challenges in 3D reconstruction robustness due to consumer-grade device limitations.
• Imaging errors from capture devices hinder accurate 3D reconstruction, and efficient feature extraction from reconstructed 3DFV remains underexplored.

Purpose of the Study:

• To develop a robust 3DFV reconstruction method with quality assessment and an adaptive feature extraction network.
• To address imaging errors and improve feature extraction efficiency in 3DFV biometrics.

Main Methods:

• Proposed Redundancy-Compensated 3DFV (RC-3DFV) reconstruction method to mitigate imaging errors.
• Unfolded 3DFVs into 3D finger texture (3DFT) and 3D finger shape (3DFS) maps for dimensionality reduction.
• Implemented a fast 3DFV Quality Assessment (3DFV-QA) mechanism for optimal reconstruction selection.
• Introduced 3DFVFSNet, coupling frequency and spatial domain convolutions for feature extraction.

Main Results:

• Experimental results demonstrate the competent performance and efficiency of the proposed RC-3DFV reconstruction and 3DFVFSNet feature extraction methods.
• The methods effectively mitigate imaging errors and capture both global and local features.

Conclusions:

• The proposed robust 3DFV reconstruction and adaptive feature extraction methods significantly advance the field of 3DFV biometrics.
• The approach enhances the reliability and efficiency of 3D finger vein authentication systems.