Appearance based modeling and learning of the human face with application to biometrics

In this dissertation we study key problems in face processing, with a focus on the applications in biometrics, including both hard biometrics (i.e. conventional face recognition) and soft biometrics, where demographical attributes (e.g. gender and age) are recognized. We categorize face processing techniques into two classes: appearance-based approaches that model facial appearance holistically, and feature-based approaches, which rely on the localization of facial feature points. In this dissertation we argue that appearance-based approaches are well suited for various face processing tasks. A fully automatic face processing system consists of several major modules: detection, alignment and recognition. In this dissertation we study the modeling and learning issues in the face alignment and recognition stages, and propose a series of algorithms to tackle these problems, demonstrating how recent developments in machine learning and applied mathematics can be employed to create effective solutions in building fully automatic, appearance-based face processing systems. Although face detection is not covered, as it is a relatively well-solved problem, some algorithms proposed in this dissertation can potentially be applied to that problem as well.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Electrical & Computer Engr

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2010

Author dc:creator

• Xu, Xun

Contributors dc:contributor

• Huang, Thomas S.
• Ahuja, Narendra
• Levinson, Stephen E.
• Liang, Zhi-Pei

• computer vision
• pattern recognition
• machine learning
• face recognition
• face alignment
• biometrics

Statement dc:rights

• Copyright 2010 Xun Xu

Language dc:language

en

Handle dc:identifier

http://hdl.handle.net/2142/16531

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/16531