Image Analysis and Pattern Recognition Techniques in Fingerprints (From International Forensic Symposium on Latent Prints -- Proceedings, 1987, Quantico, Virginia, P 61-67, -- See NCJ-113506)

• NCJ Number: 113514
• Author(s): A F Lehar
• Date Published: 1988
• Length: 7 pages
• Sale Source: Superintendent of Documents, GPO; National Institute of Justice/; NCJRS Photocopy Services
• Publication Type: Report (Technical)
• Language: English
• Country: United States of America
• Note: For microfiche, see NCJ-113506.

Annotation

This analysis of current pattern recognition techniques in fingerprints addressed approaches to binarization, measuring fingerprint quality, and the detection of minutiae.

Abstract

While automated systems have been refined considerably, they still cannot match the ability of an expert human examiner to process poor quality fingerprints. The first step in any automated fingerprint procedure is the binarization of the fingerprint. Methods discussed are thresholding algorithms, use of a spatial frequency filter, and aggressive filtering that makes direct use of the Fournier transform. The paper explains quantifying fingerprint quality by measuring local variance found in an enhanced grey-level image from which all features dissociated from ridges have been removed. Methods to detect minutiae features and subject them to a series of quantification tests are detailed. Steps toward an automatic classification system are outlined. Photographs, a table, charts, and two references.