Effect of spatial and gray level resolutions on the accuracy and processing speed of digitally-enhanced fingerprint recognition systems

Biometric recognition is increasing in popularity as opposed to older methods of recognition (passwords, identification cards, etc.) and rightfully so. They are able to allow access based on who a person is rather than what they have. One of the most popular forms of biometrics is fingerprint recognition [1]. No fingerprints of any two individuals, even identical twins, are alike. Their distinctiveness makes them very appealing as an identification tool [1]. However, even with the diversity of fingerprints, they still rely on technology to perform recognition. Fingerprint recognition is a very time consuming process and the time it takes is directly related to the systems accuracy, security, and number of users. There are cases where the time recognition takes is more important than it having near perfect accuracy and security. This thesis focuses on reducing both the spatial and gray level resolutions of fingerprint images in combination with enhancement techniques to determine how much of a reduction in resolution can be made while still providing a reliable recognition in order to improve the processing speed of the recognition process. This thesis shows that performing a Laplacian enhancement using filter separability followed by a 50% spatial-resolution reduction provides the best tradeoff between processing speed gained and accuracy lost.