An Integrated Representation of Fingerprint Patterns

The use of fingerprint features for identification purposes has a long history and is becoming increasingly popular. However, the massive size of existing fingerprint databases means that a complete manual search of the database in order to match a given fingerprint with a print in the database is out of the question, and at least part of the search process must be automated. In addition, a number of recent court challenges to the validity of fingerprint based identification has lent some urgency to the task of placing the identification process on a firm mathematical and scientific foundation. Hitherto, this task has been hampered by the lack of a standardised representation of a fingerprint pattern that encodes only the essential features of the pattern, excluding the features that are not relevant for identification purposes. Recently a representation has been proposed that considers the ridges and valleys as analogous to optical fringe patterns, with ridge terminations and bifurcations represented as “phase spirals” – points in the wave phase field where a phase singularity exists. We demonstrate how the concept of a phase field containing phase spirals, employed at two different stages in the analysis of a fingerprint image, can lead to an economical yet complete representation of the fingerprint pattern. As well as reducing the hardware requirements for image storage, it is expected that this will facilitate the task of comparing two prints to generate a match score, leading to faster and more accurate identification.