Face Recognition using Distortion Models

In this Bachelor thesis the use of distortion models in context of face recognition is evaluated. Previous work on optical character recognition indicated the usefulness of distortion models for image recognition tasks. On the other hand it has been shown that local appearance-based features such as DCT-features can be used to accommodate for illumination variances. In this thesis, these two approaches are combined to a face recognition framework that demonstrates robustness against variances in illumination facial expressions and registration errors. These qualities are examined on two databases.