Abstract
Today's legal restrictions that protect the privacy of biometric data arehampering fingerprint recognition researches. For instance, all publicdatabases of high-resolution fingerprints ceased to be publicly available. Toaddress this problem, we present an approach to creating high-resolutionsynthetic fingerprints. We modified a state-of-the-art fingerprint generator tocreate ridge maps with sweat pores and trained a CycleGAN to transform thesemaps into realistic prints. We also create a synthetic database ofhigh-resolution fingerprints using the proposed approach to propel furtherstudies in this field without raising any legal issues. We test this databasewith two existing fingerprint matchers without adjustments to confirm therealism of the generated images. Besides, we provide a visual analysis thathighlights the quality of our results compared to the state-of-the-art.