Biometric Authentication: A Machine Learning Approach

8.2. Facial Feature Extraction Techniques

Ideal facial feature extraction techniques are able to (1) eliminate the influence of all environmental, non-identity-related variations and (2) maximize the difference between faces that belong to distinctive people. Numerous methods have been proposed that would attempt to extract such features to facilitate the recognition process. These methods can be classified into two categories: feature-invariant approaches and template-based approaches [38].

8.2.1. Feature-Invariant Approaches

This type of algorithm looks for structural features that exist even when the pose, viewpoint, or lighting conditions vary. Some prior knowledge may be used to develop rules to capture the relationship between facial features. ...