See discussions, stats, and author profiles for this publication at: [Link]

net/publication/247773759

Fingerprint Recognition System : Design & Analysis

Conference Paper · January 2011

CITATION READS
1 8,740

3 authors, including:

Dev Nath
Google LLC
6 PUBLICATIONS   23 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Dev Nath on 28 May 2014.

The user has requested enhancement of the downloaded file.

 Fingerprint Recognition System : Design & Analysis
Dibyendu Nath1 Saurav Ray2 Sumit Kumar Ghosh3
Dept. of Computer Science & Engineering,
Heritage Institute of Technology,
Kolkata, India.
{[Link], [Link], [Link]}@[Link]

Abstract— Fingerprint Recognition is one of the research (EER) of 1.31 x 10-5 for a hand-geometry-based verification
hotspots in Biometrics. It refers to the automated method of system and of 2 x 10-3 for a face-based verification system.
verifying a match between two human fingerprints. It is Similarly Pankanti et al [5] also showed that there is limited
essentially a challenging pattern recognition problem where two probability of correspondence of two fingerprints.
competing error rates: the False Accept Rate (FAR) and the
False Reject Rate (FRR) need to be minimized. Advancement of B. Classification & Indexing of Fingerprints
computing capabilities led to the development of Automated Fingerprint authentication includes two subdomains: one
Fingerprint Authentication Systems (AFIS) and this led to
extensive research especially in the last two decades. In this
is fingerprint verification (Am I who I claim I am?) and the
paper, we attempt to give a comprehensive scoping of the other is fingerprint identification (Who am I?), the latter
fingerprint recognition problem and address its major design being more difficult requiring extensive indexing and
and implementation issues as well as give an insight into its classification of fingerprints for efficient retrieval.
future prospects.

Keywords- Fingerprint Recognition, Biometrics, Identification,

Verification, Security, Authentication

I. INTRODUCTION
a. Arch b. Left Loop c. Right Loop
In order to access the Internet or any other important
resource safely, high-security authentication systems are
essential. However studies [10] show that users usually
choose weak passwords, frequently re-use passwords across
multiple sites and often forget them. According to the 2002
NTA Monitor Password Survey, heavy web users have an
average of 21 pass-words, 81% of users choose a common d. Tented Arch e. Twin Loop f. Whorl
password and 30% write their passwords down or store them Fig. 1: Fingerprint classification involving 6 classes - critical points in a
in a file. Automated identity authentication using fingerprint fingerprint called core & delta marked as circles & triangles
recognition [4, 3] is an effective solution in such cases. Nearly all fingerprint classification schemes used today
Historically speaking, fingerprints have been long are derived from the famous ―Henry System‖ [1] – a detailed
associated with criminology, specifically forensics. fingerprint indexing method for aiding manual fingerprint
Development of cheaper and robust automated fingerprint comparison. For instance, the FBI uses one variant which
authentication systems coupled with the inherent ease of recognizes eight different types of patterns: radial loop, ulnar
fingerprint acquisition, has led to its widespread commercial loop, double loop, central pocket loop, plain arch, tented
and civilian applications. One of the world’s largest arch, plain whorl, and accidental.
fingerprint recognition systems is the Integrated Automated Whorls are usually circular or spiral in shape. Arches
Fingerprint Identification System (IAFIS), maintained by the have a mound-like contour, while tented arches have a spike-
FBI in the US since 1999. like or steeple-like appearance in the center. Loops have
A. Fingerprint as a Biometric concentric hairpin or staple-shaped ridges and are described
as "radial" or "ulnar" to denote their slopes; ulnar loops slope
“Two like fingerprints would be found only once every 1048 toward the little finger side of the hand, radial loops toward
years” — Scientific American, 1911. the thumb.
Individuality of fingerprints is based on empirical Fingerprint classification & indexing is a difficult pattern
observations. However Golfarelli et al [6] formulated the recognition problem due to small inter-class variability
optimum Bayesian decision criterion for a biometric compared to large intra-class variations in fingerprint
verification system and obtained a theoretical equal error rate patterns. Germain et al [15] describe a popular efficient
technique for indexing into large fingerprint databases using An important breakthrough in sensor technology was the
minutiae triplets in their indexing procedure. More efficient development of optical sensors based on fiber-optics as
classification schemes have also been proposed like [7] by described in the US patent [21], leading to sensor
Jain, et al. miniaturization and enhanced portability.

II. FINGERPRINT FEATURES

A fingerprint is an impression of the epidermal ridges of a
human fingertip. A hierarchy of three levels of features,
namely, Level 1 (pattern), Level 2 (minutiae points) and
Level 3 (pores and ridge shape) are used for recognition
purposes. Most AFISs employ Level 1 & Level 2 features.
Level 1 features refer to the overall pattern shape of the
unknown fingerprint—a whorl, loop or some other pattern.
This level of detail cannot be used to individualize, but it can
help narrow down the search. Level 2 features refers to
specific friction ridge paths — overall flow of the friction [Link] Solid-State Sensor [20]
ridges and major ridge path deviations (ridge characteristics
called minutiae) like ridge endings, lakes, islands,
bifurcations, scars, incipient ridges, and flexion creases.

a. Level 1 b. Level 2

b. Optical Sensor using FTIR

Fig. 3: Fingerprint Sensors
d. Fingerprint Minutiae
c. Level 3
Solid-state touch and sweep sensors — silicon-based
Fig. 2: Fingerprint Features devices that measure the differences in physical properties
such as capacitance or conductance of the friction ridges and
valleys dominate in commercial applications. Tartagni and
Level 3 detail [14] refers to the intrinsic detail present in a Guerrieri [22] describe a feedback capacitive sensing scheme
developed fingerprint — pores, ridge units, edge detail, scars using a 200x200 element sensor array implement in standard
etc. High resolution sensors (∼1000dpi) are required for 2-metal CMOS technology. Jeong-Woo Lee et al [20]
extraction of Level 3 features. But as [8] shows, EER values discusses another such solid-state sensor, based on capacitive
are reduced (relatively ∼20%) using them along with Level 1 differences, capable of producing 600dpi fingerprints. Many
& 2 features. Moreover Level 3 features offer greater success commercially available sweep sensors like Fujitsu MBF320
in partial fingerprint recognition as shown in [9]. are based on such low-power solid-state devices.
A special case of off-line
III. FINGERPRINT SENSING
sensing is the acquisition of a
latent fingerprint from a crime
Fingerprint sensing techniques can be of two types – off-
scene [19]. Used extensively in
line scanning and live-scanning. In off-line sensing
forensics, latent prints are
fingerprints are obtained on paper by “ink technique” which
accidental impressions left by
are then scanned using paper scanners to produce the digital
friction ridge skin on a surface,
image. Most AFISs use live-scanning where the prints are
due to natural secretions of the
directly obtained using an electronic fingerprint scanner.
eccrine glands present on skin.
Almost all the existing sensors belong to one of the three Fig. 4: Latent Fingerprint
While tremendous progress has
families: optical, solid-state, and ultrasound.
been made in plain fingerprint matching, latent fingerprint
Optical sensors, based on the frustrated total internal
matching continues to be a difficult problem. Poor quality of
reflection (FTIR) technique are commonly used to capture
ridge impressions, small finger area, and large non-linear
live-scan fingerprints in forensic and government
distortion are the main difficulties in latent fingerprint
applications. They are the most common fingerprint sensors.
matching, compared to plain fingerprint matching.
 IV. FEATURE EXTRACTION TECHNIQUES distance transform method to obtain a near-linear time
binarization of fingerprint images.
For the purpose of automation, a suitable representation Fingerprint ridge thinning is basically elimination of
i.e. feature extraction of fingerprints is essential. This redundant pixels till each ridge is just one pixel thick. An
representation should have the following properties – innovative iterative thinning technique has been proposed by
 Retention of discriminating power of each Ahmed and Ward [32] while a multi-scale thinning approach
fingerprint at several levels of resolution has been proposed by You, et al [33].
 Easy computability
 Amenable to automated matching algorithms
 Stable and invariant to noise and distortions
 Efficient and compact representation
Several feature extraction methods have been proposed
and implemented successfully over the years. Roughly
speaking there are four categories of methods based on
fingerprint feature extraction by image processing [11].The
first category of methods extract minutiae directly from the
gray-level image [1, 23, 25, 34] without using binarization
and thinning processes while the second category extracts
features from binary image profile patterns [15, 25, 26]. The
third category of methods uses machine learning [25, 28, 29] [Link] after marking b. Real Minutiae after false removal
for extracting minutiae and the last category extracts
Fig. 5: Minutiae Extraction
minutiae from binary skeletons [2, 30].
After initial fingerprint feature extraction some post-
processing is required for removing false or spurious
minutiae detected in highly corrupted regions or introduced
by previous processing steps (e.g., thinning). Chen and Kuo
Binarization
[24] proposed a three-step false minutiae filtering method,
which dropped minutiae with short ridges, minutiae in noise
regions, and minutiae in ridge breaks using ridge direction
information. Another method for removing all the spurious
pixels generated at the thinning stage in order to facilitate
a. Original Gray-scale Image b. Binarized Image subsequent minutiae filtering has been proposed by Zhao and
Tang [30].

V. FINGERPRINT MATCHING TECHNIQUES

Matching fingerprint images is an extremely difficult

Ridge
problem, mainly due to the large variability in different
Thinning impressions of the same finger (i.e., large intra-class
variations). Fingerprint matching algorithms are roughly
classified into 3 major categories –
C. Correlation-based Matching :
Two fingerprint images are superimposed and the
c. Orientation Field d. Image after Ridge Thinning
correlation between corresponding pixels is computed for
Fig. 4: Minutiae Extraction from Fingerprint Image different alignments (e.g. various displacements and
rotations). Fourier transform [12] as well as Fourier-Mellin
Binarization is the process by which an enhanced gray- Transform [13] can be used to speed up the correlation
level image is transformed into a binary image for subsequent computation.
feature detection. Good binarization algorithms should
minimize information loss and also provide efficient D. Feature-based (or Minutiae- based) Matching :
computational complexity. A binarization approach based on Typical fingerprint recognition methods employ feature-
the peak detection in the cross section gray-level profiles based matching, where minutiae (i.e., ridge ending and ridge
orthogonal to the local ridge orientation has been proposed bifurcation) are extracted from the registered fingerprint
by Ratha, et al [31]. Liang et al [27] proposed an Euclidean image and the input fingerprint image, and the number of
corresponding minutiae pairings between the two images is Latent fingerprint matching poses another whole new set
used to recognize a valid fingerprint image. Alternatively, of problems altogether. Compared to good quality full
Jain et al. [2] used a string matching technique while Isenor fingerprints acquired using live-scan or inking methods
and Zaky [17] propose a graph-based fingerprint matching during enrollment, latent fingerprints are often smudgy and
algorithm. Fan et al. [18] describes a fingerprint verification blurred, capture only a small finger area, and have large
algorithm based on a bipartite graph construction between nonlinear distortion. Hence they require enhanced extraction
model and query fingerprint feature clusters. and matching techniques to make latent fingerprint
The minutiae matching problem has been generally recognition free of manual matching and fully automated.
addressed as a point pattern matching problem which has
been extensively studied yielding families of approaches VIII. CONCLUSIONS
known as relaxation methods, algebraic and operational
research solutions, tree-pruning approaches, energy- Fingerprint Authentication has been studied for well over
minimization methods, Hough transform, etc. a century. However, its use has truly become widespread and
mainstream only in the last few decades due to development
E. Pattern-based (or Image-based) Matching of automated fingerprint recognition systems. The ever-
Pattern based algorithms compare the basic fingerprint increasing demand for reducing the error and failure rates of
patterns (e.g., local orientation and frequency, ridge shape, automated fingerprint recognition systems and the need for
texture information) between a previously stored template enhancing their security have opened many interesting and
and a candidate fingerprint. The images need to be aligned in unique research opportunities that encompass multiple
the same position, about a central point on each image. The domains such as image processing, computer vision,
candidate fingerprint image is then graphically compared statistical modeling, cryptography, and sensor development.
with the template to determine the degree of match. Our preliminary analysis shows that fingerprints have been
The image-based techniques include both optical as well proven to be an excellent if not the best biometric and its
as computer-based image correlation techniques. Recently, potential has not yet been fully realized.
several transform-based techniques have also been explored. But still, issues such as fingerprint authentication at a
For instance, a phase-based fingerprint image matching distance, real-time identification in large-scale applications
technique using 2D discrete Fourier transforms has been with billions of fingerprint records, developing secure and
proposed by Ito, et al [35] while Hamamoto [16] describes a revocable fingerprint templates that preserve accuracy, and
Gabor filter based fingerprint matching technique. scientifically establishing the uniqueness of fingerprints will
likely remain as grand challenges in the near future.
VII. MAJOR IMPLEMENTATION & DESIGN ISSUES

A fingerprint recognition system can make two types of REFERENCES

errors: a false match, when a match occurs between images
from two different fingers, and a false non-match, when [1] E. Henry, Classification and Uses of Finger Prints, Routledge,
London, 1900.
images from the same finger are not a match. Thus the chief [2] A. K. Jain, L. Hong, and R. M. Bolle, ―On-line fingerprint
objective behind the design of a good fingerprint matching verification‖, IEEE Trans. on Pattern Analysis and Machine
system is to reduce both these errors. However both the error Intelligence, 19(4):302–313, April 1997.
rates cannot be reduced simultaneously as they are inversely [3] D. Maltoni, D. Maio, A. K. Jain & S. Prabhakar, Handbook of
Fingerprint Recognition, Springer, 2003.
dependent on each other. [4] P. Komarinski, Automated Fingerprint Identification Systems, Elsevier
Another important design issue is the security of the Academic Press, 2004
fingerprint recognition system itself along with the [5] S. Pankanti, S. Prabhakar, and A. K. Jain, ―On the Individuality of
fingerprint template database. The unauthorized use or Fingerprints‖, IEEE Transactions on PAMI, Vol. 24, No. 8, pp. 1010-
1025, 2002.
disclosure of fingerprint template information from such [6] Golfarelli M., Maio D., and Maltoni D., ―On the Error-Reject Tradeoff
databases can be a serious security and privacy threat. in Biometric Verification Systems‖ IEEE Transactions on Pattern
Although fingerprint recognition has been extensively Analysis and Machine Intelligence, vol. 19, no.7, pp. 786-796,1997.
studied, there are still many open research problems in this [7] A. K. Jain, S. Prabhakar and S. Pankanti, ―Matching and Classification:
A Case Study in Fingerprint Domain‖, Proc. INSA-A (Indian National
domain, for instance : Science Academy), Vol. 67, A, No. 2, pp. 223-241, March 2001.
 Efficient Automated Fingerprint Classification [8] Anil Jain, Yi Chen, and Meltem Demirkus, ―Pores and ridges:
fingerprint matching using level 3 features,‖ 18th International
 Fully Automated Latent Fingerprint Recognition Conference on Pattern Recognition, pp. 477 – 480, 2006.
 Altered or Fake Fingerprint Detection [9] K. Kryszczuk, A. Drygajlo, and P. Morier, ―Extraction of level 2 and
 Efficient Compression of Fingerprint Templates level 3 features for fragmentary fingerprints,‖ Proc. of the 2nd
COST275 Workshop, Vigo, Spain, pp. 83-88, 2004.
 Automated Artificial Fingerprint Generation [10] D. Florencio and C. Herley, ―A large-scale study of web password
habits,‖ Proceedings of the 16th International conference on the World
Wide Web, 2007.
[11] R. C. Gonzalez and R. E. Woods., Digital Image Processing, Prentice [35] K. Ito, T. Aoki, H. Nakajima, K. Kobayashi and T. Higuchi, ―A
Hall, Upper Saddle River, NJ, 2002. fingerprint recognition algorithm using phase-based image matching
[12] Coetzee L. and Botha E.C., ―Fingerprint recognition in low quality for low-quality fingerprints‖, Proc. IEEE International Conference on
images,‖ Pattern Recognition, vol. 26, no. 10, pp. 1441-1460, 1993. Image Processing, 2005.
[13] Sujan V.A. and Mulqueen M.P., ―Fingerprint identification using space
invariant transforms,‖ Pattern Recognition Letters, vol. 23," no. 5, pp.
609-619, 2002.
[14] Q. Zhao, A. K. Jain, ―On the utility of extended fingerprint features: a
study on pores,‖ IEEE Computer Society Workshop on Biometrics,
CVPR2010, San Francisco, U.S., June 18, 2010.
[15] R. S. Germain, A. Califano, and S. Colville, ―Fingerprint matching
using transformation parameter clustering‖, IEEE Computational
Science and Engineering, pages 42–49, Oct-Dec 1997.
[16] Y. Hamamoto, ―A Gabor filter-based method for identification‖,
Intelligent Biometric Techniques In Fingerprint And Face Recognition,
pages 137–151. CRC Press, Boca Raton, 1999.
[17] D. K. Isenor and S. G. Zaky, ―Fingerprint identification using graph
matching,‖ Pattern Recognition, 19(2):113–122, 1986.
[18] K.-C. Fan, C.-W. Liu, and Y.-K. Wang, ―A fuzzy bipartite weighted
graph matching approach to fingerprint verification,‖ In Proc. of the
IEEE International Conf. on Systems, Man and Cybernetics, pages
729–733, Oct 1998.
[19] Colins M.W., ―Realizing the Full Value of Latent Prints‖, California
Identification Digest, 1992.
[20] J.-W. Lee, D.-J. Min, J. Kim, and W. Kim, ―A 600-dpi capacitive
fingerprint sensor chip and image-synthesis technique,‖ IEEE Journal
of Solid-State Circuits, Vol. 34, No. 4, April 1999
[21] Ichiro Fujieda, Yuzo Ono, Seijin Sugama, ―Fingerprint image input
device having an image sensor with openings‖, United States Patent
5446290, August 29, 1995.
[22] M. Tartagni and R. Guerrieri, ―A fingerprint sensor based on feedback
capacitive sensing scheme,‖ IEEE Journal of Solid-State Circuits, Vol.
33 Issue: 1, pages 133 - 142, Jan 1998
[23] L. Jinxiang, H. Zhongyang, and C. Kap Luk, ―Direct minutiae
extraction from gray-level fingerprint image by relationship
examination,‖ In International Conference on Image Processing (ICIP),
volume 2, pages 427–430 vol.2, 2000.
[24] Z. Chen and C. H. Kuo. ―A topology-based matching algorithm for
fingerprint authentication.‖ In IEEE International Carnahan Conference
on Security Technology, pages 84–87, 1991.
[25] B. Bir and T. Xuejun. ―Fingerprint indexing based on novel features of
minutiae triplets‖. Pattern Analysis and Machine Intelligence, IEEE
Transactions on, 25(5):616–622, 2003.
[26] C. Wu, Z. Shi, and V. Govindaraju, ―Fingerprint image enhancement
method using directional median filter,‖ In Biometric Technology for
Human Identification, SPIE, vol. 5404, pages 66–75, 2004.
[27] A. B. Xuefeng Liang and T. Asano, ―A near-linear time algorithm for
binarization of fingerprint images using distance transform,‖ In
Combinatorial Image Analysis, pages 197–208, 2004.
[28] S. Prabhakar, A. K. Jain & S. Pankanti, ―Learning fingerprint minutiae
location and type, Pattern Recognition,‖ 36(8):1847–1857, 2003.
[29] V. K. Sagar, D. B. L. Ngo, and K. C. K. Foo, ―Fuzzy feature selection
for fingerprint identification‖, In Security Technology, 1995.
Proceedings, IEEE 29th Annual 1995 International Carnahan
Conference on, pages 85–90, 1995.
[30] F. Zhao and X. Tang, ―Preprocessing and post-processing for skeleton-
based fingerprint minutiae extraction,‖ Pattern Recognition,
40(4):1270–1281, 2007.
[31] N. K. Ratha, S. Chen, and A. K. Jain, ―Adaptive flow orientation-based
feature extraction in fingerprint images,‖ Pattern Recognition, 28(11):
1657–1672, 1995.
[32] M. Ahmed and R. Ward, A rotation invariant rule-based thinning
algorithm for character recognition, Pattern Analysis and Machine
Intelligence, IEEE Transactions on, 24(12):1672–1678, 2002.
[33] X. You, B. Fang, V. Y. Y. Tang, and J. Huang ―Multiscale approach
for thinning ridges of fingerprint.‖ In Second Iberian Conf. on Pattern
Recognition and Image Analysis, vol. LNCS 3523, pp 505–512, 2005.
[34] D. Maio and D. Maltoni, ―Neural network based minutiae filtering in
fingerprints,‖ Fourteenth International Conf. Pattern Recognition,
volume 2, pages 1654–1658, 1998.

View publication stats