I.J.

Image, Graphics an
nd Signal Prrocessing, 20012, 3, 8-14
Published Onnline April 20012 in MECS (http://www.m
( mecs-press.org
g/)
DOI: 10.5815/ijigsp.2012..03.02

A Noveel Algorithm
m for Minut
M tiae M
Matchin
ng
Om Preeti Chau
urasia
Amity School of Engineeering and Tech hnology, Amitty University
Noida, Indiaa
E-mail:
E preeti.pprincy.chauraasia@gmail.co
om

Sauumya Ranjan Giri

IL&F S Technologies Ltd.
Bhhubaneswar, In
ndia
E-mail: sauumya.giri07@
@gmail.com

Anchal Garg
g
Amity School of Engineeering and Techhnology, Amitty University
Noida, Indiaa
E-maiil: agarg@amiity.edu

Abstract---- T
This paper presents
p a sim
mple and novvel where
w it bifurrcates is callled bifurcation. Minutiaee
algorithm ffor minutiaee matching in fingerpriint co
onsist of theese two basiic types, rid dge end andd
images. After correct detection
d of all
a minutiae in biifurcation. Thee overall fingeerprint recogn
nition consistss
two fingerp print images,, the algoritthm iterativeely off three steps:-
processes eaach minutiae point from two t images annd
tries to findd out the num mber of com mmon points on a. Image Preprocessinng
the basis off structural similarity
s am
mong them. W We b. Minutiae detection aand Feature ex
xtraction
try to find aall matching pairs of min nutiae betweeen c. Minutiae matching
two fingerprint images with reference to a pair of
chosen refeerence pointt. Once all the comm mon
minutiae pooints identifieed, the matching score ccan Image prepprocessing is tthe first step of fingerprintt
be calculated d using varioous existing fo
ormulas. matching.
m It is done to make image clearer too
unnderstand. In this, three ssteps are involved, Imagee
Index Terms-----Fingerprin
nt image; teemplate imagge; Quality
Q Enhancement, Binaarization and Thinning. Inn
input image; ridge end; bifurcation;
b minutiae
m im
mage quality enhancementt phase the quality q of thee
giiven image is improved so tthat the imagee will be goodd
en
nough for furtther processinng. The binarization of thee
I. INT
TRODUCTIO
ON grray scale imagge converts itt to black and white image..
A threshold is selected and applied to eaach pixel. Thee
Biometriccs is the field of science which involvves piixels above thhreshold are assigned blacck and below w
recognizing humans automatically
a via uniqque th
hreshold are assigned whitee. In thinning,, the width off
characteristiccs. Recognizinng humans by y means of thheir eaach ridge is reduced
r to onne pixel by using
u variouss
fingerprint impression is i the mostt widely ussed methods.
m Centrral line thinninng is advised.
biometrics tooday. The finggerprint recoggnition startedd in Minutiae detection
d andd feature ex xtraction stepp
1600s, but tthe use of biiometrics is expanding veery in
nvolves refinin ng of the thin inned image, detecting thee
rapidly. Finggerprints have been used wiidely in laptopps, minutiae
m pointts and then extracting features
f from
m
computers, enterprises, to college buildings eetc. im
mage. In Deteection, the w whole image is i scanned too
Fingerprint rrecognition iss used as a means
m to restrrict deetect the ridge ends and bbifurcations. In I Extraction,,
the unauthoriized users fromm accessing useful
u data. th
he coordinates, the angle off orientation an nd the type off
Fingerpriint contains complex
c pattterns of strip es, minutiae
m is com
mputed for eacch minutiae po oint.
called ridges. There exists some gap bettween the ridgges, Minutiae maatching, the thhird step invollves matchingg
called valleyys. A ridge cann spread furth
her in two wayys, th
he template im mage with thhe input image. Templatee
either it ends or bifurcatees into two rid dges. The plaace im
mage is collected at the tim me of enrollmeent and storedd
where ridge ends is calledd termination oro ridge end aand

Copyright © 2012 MECS I.J. Image,

I Graphics and Signall Processing, 2012, 3, 8-144
 A Novel Algorrithm for Min
nutiae Matchin
ng 9

in the databbase. During recognition phase,

p the inpput Two polygo ons can be ccompared forr equality byy
image is com mpared againsst template im
mage. This phaase simply considering the length
th of sides and
d angles madee
decides wheether the two images are from the sam me by
y two sides.
finger or from
m different finngers.

We are ggoing to propoose a novel allgorithm for tthe

last step i.e. minutiae mattching. The baasic idea behiind
our algorithhm is to draaw all possib ble number of
polygons by joining all detected
d minu
utiae points off a
given image.. This is donee for both the template imaage
and the inpuut image. Thenn the polygon ns of one imaage
are overlappped on other image such that maximuum
numbers of ppolygons coincide.

Fig 3: pollygon-1

Fig 1: Template image

Fig 4: pollygon-2

We
W can say po olygon-1 and polgon-2 aree same if thee
fo
ollowing conditions are truee.

A = PQ, BC = QR, CD = RS, DE = ST

AB T, EF = TU,
AF
A = PU
and
a
a = p, b = q, c = r, d = s, e = tt, f = u

This is all about

a our alggorithm and we
w match twoo
fin
ngerprint sammples on thiis idea. Two o fingerprintss
match
m only when they forrm similar po olygons afterr
jo
oining the minutiae points.
Fig 2: Input
I image

Copyright © 2012 MECS I.J. Image,

I Graphics and Signall Processing, 2012, 3, 8-144
10 A Novel Algorrithm for Min
nutiae Matchin
ng

II. PRRIVIOUS WO ORK DONE ON O MINUTIA AE finngerprints aree considered matching, if their trianglee
M
MATCHING structures aree similar aaccording th he neighborr
A lot of work has beeen done succcessfully in thhis reelationship.
area in past. One of the most
m widely used
u methodss is
[3,1] to creaate a matrix, called rotatte value, of tthe Shi et al. [10] introducced two glob bal statisticall
orientation aangle differennce between each templaate feeatures of fing
gerprint imagee, including th
he mean ridgee
minutiae, Tk (1≤k≤NT), and a each inpput minutiae, Im width
w and the normalized quality estim mation of thee
(1≤m≤NI). HHere, NT and NI N represent the
t total numbber whole
w image, and propoosed a noveel fingerprintt
of minutiae iin the template and input seets, respectiveely.
The value at rotate values (k,m) represents tthe matching
m algorrithm based oon minutiae sets combinedd
difference beetween the orientation anglees of Tk and Imm. with
w the globall statistical feeatures. This algorithm
a hass
addvantage of both local and global features inn
finngerprint matching.

Xuefeng Liaang and Tetsuuo Asano [11 1] proposed a

minutia
m polygons algorithm thahat is used to match
m distortedd
fin
ngerprints. A minutia polyggon describes not only thee
minutia
m type and
d orientation buut also the minu
utia shape. Thiss
technique emplo oys an improvved distortion model
m using a
The abovve formula is used to calcculate the raddial Multi-quadric
M baasis function with parameteers. Adjustablee
distance, raddial angle andd the orientatiion for templaate paarameters makee this model m more suitable for fingerprintt
image and tthe input im mage. Then th hese values aare distortion.
compared am mong the minnutiae points, one point froom
template and one from input imagee, with certaain
threshold vallue to find thee matching poiints.
III. PROPOSE ED ALGORIT THM
For our pro oposed algoritithm we need
d to store thee
Another mmethod calledd focal point [5 5] consideredd as
co
oordinates of the detected minutiae andd the types off
reference poiint is chosen that
t is designeed as a centerr of
minutiae.
m
moment in polar coorrdinate for the fingerprrint
matching method. The focal f point for
fo any type of TABLE I. THE
E EXTRACTED FFORMAT CAN BE
B SEEN AS
fingerprint immages can be identified and a tested w with FOLLO OWS
various orienntation, elasticc distortion and noise. Absolute co ordinate
o Minutia
ae Type
79,102
2 Bifurccation
Another aapproach for minutiae
m matcching is done by
75,214
4 Ridgee End
calculating thhe differencess of spatial an
nd orientation of
64,110
0 Bifurccation
minutiae [6]..
157,18
82 Ridgee End
Fingerprinnt minutiae can be matched by using booth
the local annd global struuctures of miinutiae [7]. TThe
Thhese details are
a stored for the template and the inputt
local structuure of a minuutia describes a rotation aand
im
mage. From th hese two sets of attributes our algorithm m
translation iinvariant featture of the minutia in its
deecides whetheer the imagess are from saame finger orr
neighborhood. It is used tot find the corrrespondence of
from different finger.
f The deetail about thee algorithm iss
two minutiaee sets and inncrease the reeliability of tthe
ex
xplained in thee next sectionn.
global matching. The global
g structu
ure of minutiiae
reliably deterrmines the uniiqueness of fin
ngerprint. IV. ILL
LUSTRATION N OF THE PR ROPOSED
ALG GORITHM
According to He et al.’’s [8] matchin
ng algorithm uuse Let us say the
t total no. oof minutiae in
n the templatee
of a variablee sized boundding box makees the matchiing is 5 and the totaal number of m
minutiae in th
he input imagee
algorithm m more robust to non-linear deformatiion is 8. The algorithm works as follows.
between finggerprint imagees.

Another interesting teechnique wass proposed [ 9],

where the m minutiae poiints are conn nected using a
Delaunay trriangulation and analyzees the relatiive
position and orientation off each minutiaa with respectt to
its neighborss obtained byy the triangle structure. Tw wo

Copyright © 2012 MECS I.J. Image,

I Graphics and Signall Processing, 2012, 3, 8-144
 A Novel Algorrithm for Min
nutiae Matchin
ng 111

TABLE II. FOR TEMPLATE

E TABLE
T V. FOR IINPUT IMAGE
Name Co ordinate Radial Minutiae Tyype P Q R S T U V W
Distance Q --
A 79,102 0 Bifurcationn
R 71° --
-
B 75,214 112 Ridge end
S 120° 48°
4 --
C 64,110 17 Bifurcationn
T 128° 56°
5 8° --
D 157,182 111 Ridge end
U 87° 16°
1 32° 40° --
E 50,255 156 Bifurcationn
V 23° 48°
4 97° 105° 64° --
W 58° 13°
1 61° 69° 29° 35° --
TABLE III. FO
OR INPUT IMAG
GE
Name C
Co ordinate Radial Minutiae Typee Here
H the angle 71° represennts the angle RPQ
R = QPR= =
Distance
P 75,103 0 Bifurcation 711°. ‘P’ is thee chosen refeerence point. So the anglee
Q 87,90 18 Bifurcation reepresents the angle made bbetween ‘Q’ anda ‘R’ madee
R 172,146 106 Ridge end with
w the referen nce point ‘P’.
S 108,206 108 Ridge end
T 101,254 153 Bifurcation
U 224,228 194 Ridge end
V 187,52 123 Ridge end
W 231,133 159 Ridge end

The radial ddistance is caalculated using

g the followiing
formula.

The radiaal distance is the

t distance of
o the coordinaate
from a choseen reference point. So if thee reference pooint Fiig 4: Figure show
ws the arrangemennt of minutiae poiints on the image

and the cooordinate are the same, then t the raddial

Th
he following table
t shows thhe result of matching of thee
distance will be zero.
minutiae
m pointss on the basis oof radial distaance only.
During ppairing the minutiae
m we must
m be care ful
about the tyype of minutiiae (Ridge an nd Bifurcationn).
Bifurcations should be paired with bifu urcation of othher
image and tthe same is for ridges. Let L us pair tthe TABLE VI.
V RADIAL DIST
TANCE COMPA
ARISON
reference pooint ‘A’ of template im mage with tthe Minutie
M Radial iae
Minutia Radial Difference
reference pooint ‘P’ of input imag ge. Now othher Points of Distance Points of Distance in radial
Teemplate fro
om Inputt from distance
calculations aare as followss. Refer
rence imagee Reference
po
oint point
A 0 P 0 0
Calculatioon of Anglees made bettween minutiiae B 112 S 108 4
points with thhe reference point
p as centerr. Angles can be C 17 Q 18 1
easily calculaated from the coordinates of
o three points . D 111 R 106 5
E 15
56 T 153 3
TABLE IV. FOR
F TEMPLATE
E
A B C D E
B -- 60° 46° 9°
C 600° -- 106° 51° If we set the threshold vaalue of difference betweenn
D 466° 106°° -- 55° th
he radial distaances is 5, theen all the abo ove points aree
E 9° 51° 55° -- matched.
m So the
t minutiae points paired d (Only radiall
diistance is conssidered) till noow are as follo
ows:

Copyright © 2012 MECS I.J. Image,

I Graphics and Signall Processing, 2012, 3, 8-144
12 A Novel Algorrithm for Min
nutiae Matchin
ng

A–P his implies thee following paairs are correcct.

Th
B–S A–P
C–Q D–R
D–R B–S
E–T E–T

Now let us compare the angles madee by each pooint So we can conclude
c that the given tem
mplate and thee
with other pooints. nput image have only 4 minnutiae points matched withh
in
‘A
A’ as reference point in thee template imaage and ‘P’ ass
reeference point in the input im
mage.
T
TABLE VII. ANG
GLE COMPARIS
SON
A/P B/SS C/Q
Q D/R E/T The abovee matched m minutiae poiints can bee
B/S --
diiagrammaticallly viewed as follows.
C/Q 60°°/120° --
D/R 46°°/48° 1066°/71° --
E/T 9°//8° 51°//128° 55°/56° --

Let us diiscuss in detaail about the above table. It

shows the anngle CAB = 60°
6 and the an ngle QPS= 1200°.
But we havve the followiing pairs from the previoous
analysis.

A–P
C–Q
B–S Fig
F 6: Figure show
ws the arrangemen
ent of all matched
d minutiae points
on the templlate image
If this is tthe scenario, then
t the anglees must be sam
me.
So this contrradicts and it means
m ‘C’ maay be a differeent
minutiae poiint and not the t counterpaart of ‘Q’. T The
same condittion is for ‘B B’ and ‘S’. The differennce
between the angles is 1200° - 60° = 60°°. Let us say tthe
threshold vallue for angle difference
d is 5°,
5 and then thhis
does not satissfy the criteriaa.

Fig
F 7: Figure show
ws the arrangemen
ent of all matched
d minutiae points
on the inpuut image

We can see that all thee 4 minutiaee points form m

similar structures in both thee template and
d input image.
Fig 5: Figuree shows the arranngement of minuttiae points on the
templlate image and thhe input image resspectively
The same procedure
p is repeated for each chosenn
paair of refereence points. The highestt number off
So if we calculate the diifference betw ween angles w we
matching
m point is the result.
find only thee following anngles lie belo
ow the threshoold
value.
V. TH
HE ALGORIT
THM FOR MIINUTIAE
Angle DAB and angle RPS, difference isi 2°. MAATCHING
Angle EAB aand angle TPS
S, difference is 1°.
Angle EAD aand angle TPR
R, difference is
i 1°.
1. Repeat step-2 to step-5 until all minutiaee
points consumed.

Copyright © 2012 MECS I.J. Image,

I Graphics and Signall Processing, 2012, 3, 8-144
 A Novel Algorithm for Minutiae Matching
2. Take one minutiae point from template and
one from input image as reference and repeat
step-3 to step-4 until all minutiae points are
consumed.
3. Calculate radial distances for all points from
the reference point for both the template and
the input image.
4. Pair one minutiae point from template with
one from input image if the difference in
radial distance of the points ≤ certain
threshold value and their type are same.
a. If the above condition is true, then find
the angle between two minutiae of an
image made with the reference point.
Find the same angle in other image.
b. If difference between the angles ≤ certain
threshold value, then pair minutiae points
from template and input image.
5. Count the number of pairs (non repeating
pairs) and this represents total number of
matching minutiae points for a particular pair
of reference points. Store the count value.
6. Finally select the maximum count value. This
is the final matching minutiae points.
VI. CONCLUSION AND FUTURE WORK
We have developed an algorithm for minutiae
matching on the basis of structural similarity of
minutiae points on the finger print image. There are
only two kinds of minutiae that are considered in the
above algorithm, in fact various types of minutiae can
be matched using the same algorithm. We have tested
the algorithm and found it efficient to a great extent.
The algorithm is tad slower than some other algorithm
but gives better result. In future the time complexity of
the algorithm can be reduced and the same algorithm
can be used in applications other than fingerprint
matching where 2D structural similarity is concerned.
REFERENCES
[1] Graig T. Diefenderfer, June 2006. Thesis on
“Fingerprint Recognition” at Naval Postgraduate
School, Monterey, California,.
[2] Ravi J, K. B. Raja, Venugopal K. R., Fingerprint
Recognition using Minutiae Score Matching,
International Journal of Engineering Science and
Technology Vol.1(2), 2009, 35-42.
Copyright © 2012 MECS
13
[3] Xiping Luo, Jie Tian and Yan Wu, A Minutia
Matching Algorithm in Fingerprint Verification,
[4] P. Kumar, S. R. Giri, G. R. Hegde and K. Verma,
A Novel Algorithm to Extract Connected
Components in a Binary Image of Vehicle
License Plates, IJECCT 2012, Vol. 2 (2), Page
27-32.
[5] Krisakorn Rerkrai and Vutipong Areekul, “A
New Referance Point For Fingerprint
Recognition”.
[6] F.A. Afsar, M. Arif and M. Hussain, Fingerprint
Identification and Verification System using
Minutiae matching, National Conference on
Emerging Technologies 2004.
[7] Xudong Jiang and Wei-Yun Yau, Fingerprint
Minutiae Matching Based on the Local And
Global Structures.
[8] Yuliang He, Jie Tian, Xiping Luo, Tanghui Zhang.
Image enhancement and minutiae matching in
fingerprint verification. Elsevier, Pattern
Recognition Letters 24 (2003) 1349–1360
[9] Giuseppe Parziale and Albert Niel. Fingerprint
Matching Using Minutiae Triangulation.
[10] Peng Shi, Jie Tian, Qi Su, Xin Yang. A Novel
Fingerprint Matching Algorithm Based on
Minutiae and Global Statistical Features.
Biometrics: Theory, Applications, and Systems,
2007. BTAS 2007. First IEEE International
Conference. 27-29 Sept. 2007
[11] Xuefeng Liang and Tetsuo Asano. Fingerprint
Matching Using Minutia Polygons. 18th
International Conference on Pattern Recognition -
Volume 01.
Om Preeti Chaurasia is pursuing M.Tech in Amity
University, Noida, India. She received the B.Tech
degree from Noida Institute of Engineering and
Technology, Noida, India in 2009. Her research areas
include Data Mining, Artificial Intelligence and
Software Engineering.
Saumya Ranjan Giri is a Software Engineer at IL&FS
Technologies Ltd., Bhubaneswar. He received the
B.Tech degree from Biju Patnaik University of
Technology, Odisha, India in 2009 and the M.Tech
degree from IIIT Bhubaneswar, Odisha, India in 2011.
He has published articles in International Journals and
Conferences. He has presented papers in International
Conferences. His current research areas include Image
Processing, Data Structures and Data Compression.
I.J. Image, Graphics and Signal Processing, 2012, 3, 8-14
14 A Novel Algorithm for Minutiae Matching

Mrs. Anchal Garg is an Assistant Professor in Amity

School of Engineering and Technology, Noida, India.
She received the MCA degree from Punjabi University
Campus, Patiala, India in 2002 and M.Tech degree
from Rajsthan Vidyapeeth, Rajasthan, India in 2006.
Her research areas include Data Mining, Operations
Management and Human Computer Interaction.

Copyright © 2012 MECS I.J. Image, Graphics and Signal Processing, 2012, 3, 8-14