Date of Award

Summer 2005

Document Type

Thesis

Degree Name

Master of Science (MS)

Department

Electrical & Computer Engineering

Program/Concentration

Electrical Engineering

Committee Director

Vijayan K. Asari

Committee Member

Stephen A. Zahorian

Committee Member

Sacharia Albin

Call Number for Print

Special Collections LD4331.E55 S256 2005

Abstract

A feature specific modular-PCA (Principal Component Analysis) approach on face images in multiple views for pose and illumination invariant face recognition is presented in this thesis. Principal components are extracted from different sub-modules of the image and are concatenated to make a single signature vector to represent a face region in a particular viewing angle. Additional principal components are extracted horn image regions representing key facial feature and are added as an extension of the signature vector. Feature specific modular-PCA approach is capable of recognizing faces in varying illumination conditions and facial expressions as the modular components represent the local information of the facial regions. Separate feature specific signature vectors are formed for sets of face images at different viewing angles to help identify a person at different poses.

Development of novel techniques for accurate location of eyes and nose and determination of pose angle of a person in a complex lighting environment is also presented in this thesis. An adaptive progressive thresholding technique is applied to spot the darkest regions representing the eyes in a face. The nose region is located by performing cumulative histogram based thresholding of the gradient image formed below the eye region. Pose angle is determined by performing a geometric analysis on the eyes and nose locations. Several experiments have been conducted on standard test databases including the face database provided by the Face Recognition Grand Challenge (FRGC) program of the United States Government.

It is observed that the proposed face recognition approach can improve the recognition rates in varying lighting, expression and pose angles. Research is underway to apply the feature specific multi-view approach with a non-linear dimensionality reduction technique suitable for representing faces in extremely complex lighting conditions.

Rights

In Copyright. URI: http://rightsstatements.org/vocab/InC/1.0/ This Item is protected by copyright and/or related rights. You are free to use this Item in any way that is permitted by the copyright and related rights legislation that applies to your use. For other uses you need to obtain permission from the rights-holder(s).

DOI

10.25777/2jk6-cy53

Download

Share

COinS