Impact of Henry classification on the entropy of fingerprint images

Chandrasekaran Vandhana, Purdue University

Abstract

Fingerprints have been accepted as a standard form of identification for decades. Passwords, tokens, and PINs despite obvious vulnerabilities are considered secrets due to the established nature of entropy associated with it. Theoretical and empirical studies are carried out to determine the amount of information present in biometric feature that uniquely identifies an individual. The empirical study carried out by Young (2007) showed the probability of minutiae occurring in the center of the image to be higher than in the edges. The study also used Shannon's Information Theory joint entropy equation to determine the entropy of fingerprint images. Studies (Chen & Jain, 2001; Young & Elliott, 2007) showed fingerprint images categorized based on the Henry system of classification and image qualities to have a statistically significant difference in minutiae distributions and counts. Because fingerprints' entropy are estimated by the underlying minutiae clustering property, the current study looked at whether different patterned fingerprint images, prints from different fingers, images of different qualities and finally fingerprints from participants of different age groups have a statistically significant difference in the amounts of entropy. Analysis showed amount of entropy in fingerprints of different patterns to differ significant statistically—the whorl pattern was identified to have the highest amount of entropy (13.81 bits) on an average followed by loops (12.52 bits) and arches (11.88 bits). The analysis was extended further to see whether the interaction, between the finger types and Henry classification, has a significant effect on the entropy of fingerprints. The results showed the interaction effect between finger type and the Henry classification to have no significant effect on the amount of fingerprint entropy. Yet, significant difference in the amount of entropy was observed among fingerprints of different fingers—with the ring finger ranking first with the highest amount of entropy (13.6 bits) on an average followed by middle (13.17 bits), index (12.68 bits), and finally little (11.60 bits) finger. The analysis was extended further to determine whether fingerprint images of different qualities have different amount of entropy. The analysis showed good (12.85 bits) and adequate (12.64 bits) quality fingerprint images on an average to have the highest amount of entropy followed by medium (11.81 bits) and poor (10.86 bits) quality images. To summarize, the current study calculated entropy by borrowing ideas from the study conducted by Ratha, Connell, and Bolle (2001) and Young (2007), looked at how the minutiae clustering tendency impacts the entropy of fingerprint images, and showed that images of good quality, whorl patterned images and fingerprints from ring finger have higher amount of entropy, on an average, compared to their counterparts.

Degree

M.S.

Advisors

Elliott, Purdue University.

Subject Area

Information science

COinS