Regularized robust coding for face recognition.

Area of Science:

• Computer Vision
• Machine Learning
• Pattern Recognition

Background:

• Sparse Representation-based Classification (SRC) is a recent method for robust face recognition (FR).
• SRC's assumption of Gaussian or Laplacian distribution for coding residuals limits its effectiveness in practical FR systems.
• The computational cost of SRC is high due to sparsity constraints on coding coefficients.

Purpose of the Study:

• To propose a novel face coding model, Regularized Robust Coding (RRC), for improved face recognition.
• To develop an efficient algorithm for solving the RRC model.
• To enhance the robustness and efficiency of face recognition systems compared to existing methods.

Main Methods:

• Developed the Regularized Robust Coding (RRC) model for robust signal regression.
• Assumed independent and identically distributed coding residuals and coefficients for a maximum a posteriori solution.
• Proposed an iteratively reweighted regularized robust coding (IR(3)C) algorithm for efficient computation.

Main Results:

• RRC demonstrates superior effectiveness and efficiency compared to state-of-the-art sparse representation methods.
• The proposed IR(3)C algorithm efficiently solves the RRC model.
• Experiments show RRC's robustness against face occlusion, corruption, lighting, and expression changes.

Conclusions:

• The RRC model offers a more effective and efficient approach to face recognition than traditional SRC.
• The IR(3)C algorithm provides an efficient solution for the RRC model.
• RRC significantly enhances face recognition robustness in challenging real-world conditions.