Maximum Correntropy Criterion for Robust Face Recognition.

Area of Science:

• Computer Vision
• Machine Learning
• Signal Processing

Background:

• Sparse representation is crucial for face recognition.
• Existing methods like l(1)-norm-based sparse representation classifier (SRC) are sensitive to noise and outliers.
• Robustness against image corruption and occlusion remains a challenge.

Purpose of the Study:

• To develop a robust sparse representation framework for face recognition.
• To improve recognition accuracy and efficiency, especially under challenging conditions.
• To create a computationally efficient algorithm less sensitive to outliers.

Main Methods:

• A sparse correntropy framework is proposed, based on the maximum correntropy criterion.
• Nonnegativity constraints are imposed on variables within the maximum correntropy criterion.
• A half-quadratic optimization technique is employed to approximate the objective function, reducing the problem to a weighted linear least squares problem iteratively.

Main Results:

• The proposed method demonstrates superior robustness against occlusion and corruption compared to state-of-the-art techniques.
• Experimental results show improvements in both face recognition accuracy and receiver operator characteristic (ROC) curves.
• The computational cost is significantly lower than traditional SRC algorithms.

Conclusions:

• The sparse correntropy framework offers a more robust and efficient solution for face recognition.
• The method effectively handles noisy and corrupted face images, outperforming existing approaches.
• This framework provides a promising direction for advancing reliable face recognition systems.