The document discusses a face recognition system that is robust to varying lighting conditions. It proposes combining illumination normalization preprocessing, local texture-based face representations like Local Binary Patterns (LBP) and Local Ternary Patterns (LTP), and distance transform-based matching. The preprocessing aims to eliminate lighting variations while preserving important appearance details. Local patterns make the representations less sensitive to noise. Experiments show the method outperforms other preprocessors across datasets and lighting conditions, providing a 88.1% verification rate at 0.1% false acceptance. The system is implemented in MATLAB.