High-Confident Block Diagonal Analysis for Multi-View Palmprint Recognition in Unrestrained Environment

Area of Science:

• Computer Science
• Biometrics
• Pattern Recognition

Background:

• Unrestrained palmprint recognition faces challenges from variable image quality, lighting, and poses in real-world scenarios.
• Existing methods often rely on subspace structures, with block diagonal properties demonstrated for palmprint data.

Purpose of the Study:

• To develop a unified learning model for robust multi-view palmprint recognition.
• To ensure a consensus block diagonal property across all views for improved feature extraction.

Main Methods:

• Proposed a novel high-confident block diagonal analysis for multi-view palmprint recognition (HCBDA MPR).
• Introduced a multi-view block diagonal regularizer to enforce a consensus block diagonal structure.
• Preserved discriminant features while learning a strict block diagonal structure across views.

Main Results:

• The proposed HCBDA MPR method demonstrated superior performance on real-world unrestrained palmprint databases.
• Achieved the highest recognition accuracies compared to existing state-of-the-art methods.
• Validated the effectiveness of the consensus block diagonal property for multi-view palmprint recognition.

Conclusions:

• HCBDA MPR offers a significant advancement in unrestrained palmprint recognition.
• The method effectively addresses challenges posed by uncontrolled environments.
• The approach provides a robust framework for identity authentication using multi-view palmprints.