Multispectral image fusion for illumination-invariant palmprint recognition

Area of Science:

• Biometrics
• Computer Vision
• Pattern Recognition

Background:

• Multispectral palmprint recognition offers high accuracy and stability for personal identification.
• Illumination variations pose a significant challenge in palmprint recognition systems.

Purpose of the Study:

• To develop a novel illumination-invariant multispectral palmprint recognition method.
• To enhance the robustness and accuracy of palmprint-based identification systems under diverse lighting conditions.

Main Methods:

• An image-level fusion framework utilizing Fast and Adaptive Bidimensional Empirical Mode Decomposition (FABEMD) and a weighted Fisher criterion.
• FABEMD decomposes multispectral images into intrinsic mode functions for illumination compensation.
• A tensor-based extreme learning machine (TELM) is employed for feature extraction and classification.

Main Results:

• The proposed fusion framework demonstrates strong robustness against illumination variations.
• High recognition accuracy achieved: 99.93% under ideal illumination and 99.50% under unsatisfied lighting.
• The TELM mechanism ensures fast learning speed and satisfying recognition accuracy.

Conclusions:

• The developed illumination-invariant multispectral palmprint recognition method is effective and robust.
• The proposed approach offers a promising solution for reliable personal identification in real-world scenarios.
• The combination of FABEMD, weighted Fisher criterion, and TELM yields superior performance.