Texture Modeling for Synthetic Fingerprint Generation

Abstract

The development of biometric recognition technologies often requires large sets of biometric data for training and evaluation purposes. The use of synthetically generated biometric samples has been explored as a means of avoiding the challenges of large scale data collection. Our paper builds on previous work in synthetic fingerprint generation research through the modeling and synthesis of texture characteristics for synthetic fingerprint generation. The proposed texture characterizing features can be modeled from real fingerprint images to generate synthetic fingerprint texture statistically representative of a particular real fingerprint database. The texture characterizing features include ridge intensity along the ridge center-lines with seven frequency components, ridge width, ridge cross-sectional slope, ridge noise, and valley noise. A comparison of these feature densities from real and synthetic fingerprints is shown, which demonstrates the effectiveness of this method of modeling and generating synthetic fingerprint textures.