Date of Award

Spring 2011

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Electrical/Computer Engineering

Committee Director

Vijayan K. Asari

Committee Member

Shirshak K. Dhali

Committee Member

Gene Hou

Committee Member

Jiang Li

Abstract

Face recognition is fundamental to a number of significant applications that include but not limited to video surveillance and content based image retrieval. Some of the challenges which make this task difficult are variations in faces due to changes in pose, illumination and deformation. This dissertation proposes a face recognition system to overcome these difficulties. We propose methods for different stages of face recognition which will make the system more robust to these variations. We propose a novel method to perform skin segmentation which is fast and able to perform well under different illumination conditions. We also propose a method to transform face images from any given lighting condition to a reference lighting condition using color constancy. Finally we propose methods to extract local features and train classifiers using these features. We developed two algorithms using these local features, modular PCA (Principal Component Analysis) and boosted tree. We present experimental results which show local features improve recognition accuracy when compared to accuracy of methods which use global features.

The boosted tree algorithm recursively learns a tree of strong classifiers by splitting the training data in to smaller sets. We apply this method to learn features on the intrapersonal and extra-personal feature space. Once trained each node of the boosted tree will be a strong classifier. We used this method with Gabor features to perform experiments on benchmark face databases. Results clearly show that the proposed method has better face recognition and verification accuracy than the traditional AdaBoost strong classifier.

DOI

10.25777/t87r-z026

ISBN

9781124719894

Share

COinS