Semester

Summer

Date of Graduation

2007

Document Type

Dissertation

Degree Type

PhD

College

Statler College of Engineering and Mineral Resources

Department

Lane Department of Computer Science and Electrical Engineering

Committee Chair

Hany H Ammar

Abstract

Automating the identification process of individuals is receiving increased attention. Fully automated image segmentation from different types of biometric images is an essential step for designing automated identification systems. In this work, we address the problem of fully automated image segmentation in the context of dental and ear biometrics. We also address the problem of evaluating the quality of segmented image by designing an automated segmentation evaluator.;For image segmentation, we first apply a mathematical morphology operator to highlight the desired objects and suppress the others, and then threshold the resulting image to separate the desired objects from the background. We next analyze the connected components obtained from the thresholded image based on their geometric properties in order to isolate the desired objects. Results on dental radiograph images show that our approach performs very well compared to the other automated approaches. In addition, it has the lowest failure rate and highest optimality, and can deal not only with the bitewing views but also with the periapical views.;In ear images, our segmentation approach achieves more than 90% accuracy based on three different sets of 3750 facial images for 376 persons.;We also present an approach for the automated evaluation of the quality of segmented images. Our approach is based on low computational-cost appearance-based features, and consists of two stages: off-line and on-line. In the off-line stage, we generate training sets by manually classifying the segmentation outcomes of proposed approach into several subclasses. Next, we create the Eigen-spaces corresponding to the different training sets. In the on-line stage, we project the outcome of segmentation onto the Eigen-spaces after view normalization, and use a Bayesian Classifier in order to determine whether the segmentation outcome is proper or improper segment. Experimental results for evaluating the segmentation outcomes of dental images and ear images indicate the benefits of the scheme.

Download

Share

COinS