The primary purpose of this study was to develop a sound, quantitative basis for assessing the quality of fingerprint images.
The primary aim of this study was to develop a quantitative basis for assessing the quality of various types of fingerprint images. Fingerprint images were obtained from several sources: a set of 117,323 anonymized images, rolled and slap prints, from 2,575 different individuals from the Federal Bureau of Investigation's (FBI's) Criminal Justice Information Services division; 516 latent prints from the NIST (National Institute of Standards and Technology) Special Database 27, as well as prints obtained in the laboratory using latent lifting methods; and a newly created database of digitally altered images of actual prints for use in determining drop-off points. The researchers tested their theoretical concepts on these actual images and then performed statistical analyses to test the validity of the results. The analyses led to the development of quantitative thresholds for unbiased selection and for the use of Level 2 detail. The researchers developed software for use in extracting minutiae, ridges, and extended feature representations of images, as well as conducted data mining in order to identify new feature types that are statistically rare in fingerprint image databases. The researchers also developed a method for conducting high-speed computing when performing the data mining work, in addition to developing an improved automated method for conducting image segmentation in which the fingerprint region is separated from the background of an image. To date, this work has improved the objective level for assessing the quality of fingerprint images, however more work is required. Implications for policy and practice are discussed. Tables, figures, and references
Date Published: May 1, 2012