| "Biometrics" are automated methods of recognizing an individual based on their physical or behavioral characteristics. Biometric recognition systems utilize the physiological or behavioral characteristics of an individual for identification. A biometric system is essentially a pattern recognition system that operates by acquiring biometric data from an individual, extracting a feature set from the acquired data, and comparing this feature set against the template set in the database. Feature extraction procedures can be made more refined by using wavelets, whose building blocks are well localized in time as well as frequency domain. This dissertation is divided into two parts: first part indulges into iris recognition methodologies, while second part presents different algorithms designed for 'liveness' detection in fingerprint scanners.; The first chapter gives a brief introduction to this thesis. The rest of the chapters are divided into two parts. The first part focuses on 'iris recognition' and is comprised of five chapters (2-6). Chapter 2 gives a brief introduction of 'iris recognition'.; In chapter 3 a novel method to perform iris recognition using biorthogonal wavelets is introduced. Effective use of biorthogonal wavelets using a lifting technique to encode the iris information is demonstrated which minimizes built in noise of iris images using in-band thresholding. Comparison with Gabor encoding, similar to the method used by Daugman and others is performed. The advantage of this algorithm over Daugman's is it may provide more flexibility for non-ideal images. For over 4,536 intra-class and 566,244 inter-class comparisons were made. FRR and FAR values of 13.6% and 0.6% using Gabor filter and 0% and 0.03% using the biorthogonal wavelets were obtained.; Chapter 4 presents effective use of Active Shape Models (ASMs) for doing iris segmentation. In practical situations it is very difficult to get iris images without any noise elements and captured without any angle deformations. A method for building flexible model by learning patterns of iris invariability from a well organized training set is demonstrated. The model specific approach taken in the work sacrifices generality, in order to accommodate better iris segmentation. (Abstract shortened by UMI.)... |
