JuCheng Yang, JinWook Shin, BungJun Min, Bin Yu, DongSun Park
Dept. of Infor.& Comm.Eng., Chonbuk National University, Jeonju, Jeonbuk, 561-756, Korea.
Keywords: Fingerprint, Recognition, FingerCode, Wavelet Transform.
Abstract: FingerCode has been an effective representation for both the local and global information in fingerprints
using their reference points. Wavelet transform is known to be a powerful tool for fingerprint enhancement
and features extraction. In this paper, a novel method for fingerprint recognition using the FingerCode in
wavelet transform domain is proposed. The proposed method includes a new reference point detection
method in sub-images of the wavelet transform. Since the proposed method can be used for both feature
extraction and pre-processing, conventional pre-processing algorithms can be eliminated from recognition
steps and hence, it lowers the overall computational complexity of the recognition. Experimental results
show that the proposed method is more accurate and reliable than a traditional FingerCode method.
Traditionally, passwords (knowledge-based security)
and badges (token-based security) have been used to
restrict access to secure systems. However, security
can be easily breached in these systems when a
password is divulged to an unauthorized user or a
badge is stolen by an impostor. The emergence of
biometrics has addressed the problems that plague
traditional identification or verification systems by
using certain physiological or behavioural
characteristics associated with an intended person,
such as fingerprints, hand geometry, iris, retina,
face, hand vein, facial thermo grams, signature, and
voice-print. Biometric indicators provide uniqueness
and have an edge over traditional security methods
in that these attributes cannot be easily stolen or
shared. Among all the biometric indicators,
fingerprints has been proven as providing one of the
highest level of reliability and extensively used in
many applications, for example, in criminal
investigations by forensic experts.
Three types of matching methods have been used
for fingerprint recognition: correlation-based
matching methods, minutiae-based matching
methods (Jang and Yau, 2000 – Liu, et al., 2000),
and texture-based (filter-based) matching methods
(Jain, et al., 1999 - 2000, Sha, et al., 2003). In
correlation-based matching methods, two fingerprint
images for matching are superimposed and the
correlation between corresponding pixels is
computed for different alignments such as various
displacements and rotations. Since the fingerprint
representation coincides with the whole fingerprint
image, these methods are quite time-consuming.
The most popular and widely used techniques for
matching are minutiae-based matching methods.
Methods in this type extract feature vectors from the
two fingerprints as sets of points in the two-
dimensional plane. They essentially consist of
finding the alignment of minutiae points between the
template and the input sets that result in the
maximum number of minutiae pairing. However, a
minutiae-based matching may not fully utilize
significant components of the rich discriminatory
information available in the fingerprints and it
usually very time-consuming (Maltoni, et al., 2003).
The texture-based matching methods use features
from the fingerprint ridge pattern such as local
orientation and frequency, ridge shape, and texture
information, which can be extracted more reliably
than minutiae points. Filter-based matching is a type
of the texture-based matching. The FingerCode
(Jain, et al., 1999 – 2000) and its improved
algorithms (Sha, et al., 2003) are shown to provide
effective representations by extracting filtered
features from the fingerprint image.
Wavelet transform have been used for fingerprint
verification and recognition recently. Selvaraj et al
Yang J., Shin J., Min B., Yu B. and Park D. (2006).
In Proceedings of the First International Conference on Computer Vision Theory and Applications, pages 161-165
DOI: 10.5220/0001369301610165
(2003) proposed a method of matching between the
input image and the stored template without
resorting to exhaustive search using both the wavelet
statistical features and wavelet co-occurrence
features. However, it uses all the pixels in the
wavelet sub-band image for the computation of the
statistical features, and it is much time-consuming.
Tico et al. (2001) suggested another matching
algorithm using wavelet domain features. They used
a feature vector of length 12 to represent a
fingerprint image. The feature vector represents an
approximation of the image energy distribution over
different scales and orientations. Fung et al. (2004)
proposed an improved approach of ref (Tico, et al.
2001). In their work, critical wavelet coefficients
were selected to form a feature vector of a
fingerprint image. However, the vector with 12
features is not sufficient to use all the information of
a fingerprint image so that the recognition rate may
not be appropriate for some applications. Lee W.K.
et al. (1997) proposed an algorithm that extracts the
dominant local orientation features in the wavelet
transform domain. The performance of the
algorithm is directly related to the accuracy of the
detection of the local directions. Mokju et al. (2004)
proposed an algorithm based on directional image
constructed using the expanded Haar Wavelet
Transform. In the work, they first obtain a
directional image, and then quantize the directional
image into a few grey-level values that represent a
range of angle orientations. In this method, the
quantizing process may be error-prone in computing
the directional information.
To overcome the drawbacks of these methods, a
new matching method is proposed in this work. We
use a sophisticated FingerCode method in the
wavelet transform domain for fingerprint
recognition. In the work, FingerCode are extracted
in the decomposed wavelet sub-band images instead
of the original fingerprint image. There are two
advantages to extract features from the wavelet sub-
band images. Since the wavelet transform is a multi-
resolution tool in signal processing, it can easily
remove the high-frequency noise, usually contained
in HH sub-band image. With this approach one can
eliminate some pre-processing steps such as noise
removing, binarization, thinning and restoration. In
addition, the size of decomposed sub-band images is
half of the original image, so that a matching method
using features extracted from sub-band images can
speed up the whole matching process comparing to
other approaches.
The paper is organized as follows: In Section 2
The theory of FingerCode is briefly reviewed. The
proposed recognition method is explained in Section
3 and its experimental results are shown in Section 4.
The conclusion remarks are given in Section 5.
The FingerCode, introduced in ref (Jain et al., 2000),
is a fixed length representation that can effectively
capture both the local and global details in a
fingerprint, with a bank of Gabor filters. The typical
FingerCode generation process can be summarized
in the following steps:
1. Locate the reference point and determine the
region of interest for a fingerprint image.
2. Tessellate the region of interest, centered at the
reference point, into a series of B (=5) concentric
bands and divide each band into k (=16) sectors.
3. Normalize each sector to a predetermined
constant mean M
(=100) and variance V
4. Filter the region of interest in eight different
directions using a bank of Gabor filters.
5. Computer the average absolute deviation from
the mean (AAD) of grey level values in each of the
80 sectors for every filtered image. The collection of
all the AAD features in each filtered image is
defined as FingerCode.
6. Rotate the features in the FingerCode cyclically
to generate five templates corresponding to five
rotations (±45
, ±22.5
, 0
) of the original fingerprint
image, thus to approximate the rotation-invariance;
7. Rotate the original fingerprint image by an
angle of 11.25
and generate its FingerCode.
Another five templates corresponding to five
rotations are generated in the same way as step 6.
8. Match the FingerCode of the input fingerprint
with each of the ten templates stored in the database
to obtain ten matching scores. The final matching
score is the minimum of the ten matching scores,
which corresponds to the best matching of the two
In this paper, we use the reference point location
method developed in ref (Sha, et al., 2003) for the
original fingerprint image, which is known to be
robust and rotation-invariance. The average
orientation of each sector is also computed for the
reference point.
The proposed algorithm for the fingerprint
recognition consists of three main steps:
1. Apply the discrete wavelet decomposition to a
fingerprint image.
2. Determine the reference point in the wavelet sub-
3. Apply the FingerCode approach to the wavelet
The first step is to apply the wavelet
decomposition. Typically, Daubechies wavelet
filters are reasonable tools for decomposing images
(Mallat, 1998), here for simplicity, Db4 is chosen as
the wavelet basis. We use Db4 wavelet to
decompose the fingerprint image into 2 levels, the
approximated sub-images LL1 and LL2 are shown
as in Fig.1. We choose the approximated sub-image
LL1 and LL2, and exclude LL3 or higher
decomposed sub-images, since the size of higher-
level decomposed sub-image is so small, and they
hardly provide unique information as FingerCode
For the approximated sub-images, the ridges and
valleys may not be clearly defined due to the
approximation as in Fig. 1(b) and (c). Hence it is
very hard to find reference points directly from these
sub-band images using the traditional method. A
new reference point detection method in the wavelet
sub-images needs to be developed and described as
the second step of the proposed method.
(a) (b) (c)
Figure: 1. (a), (b), (c) original image 101_7.tif (300×300)
and its 1-level and 2-level decomposed sub-images LL1
and LL2 (Db4 wavelets used).
The second step is to determine the reference
points from the wavelet sub-images. In here we
proposed a new method based on the method in ref
(Sha, et al., 2003). Since it is difficult to determine
the reference point in sub-band images with less
information, we first determine the reference point
(dx, dy) in the original fingerprint image using the
method in ref (Sha, et al., 2003). Then the algorithm
finds the location of the reference point in the
wavelet sub-images using the proportional location
of the reference point in the original image. Since
the decomposition of wavelet transform uses down-
sampling in half, the size of the decomposed sub-
image is a half size of the up-level image. If we find
the coordinate of the reference point according to the
left-top point (0,0) is (dx, dy) in the original image,
then the coordinates of the reference point according
to its left-top point can estimated as (dx/2, dy/2) and
(dx/4,dy/4) in the LL1 and LL2 sub-images as
shown in Fig.2 (b),(c) respectively.
The third step is using the FingerCode on the
wavelet sub-images based on the method described
in ref (Sha, et al., 2003). Since we locate the
reference point in the wavelet domain, the
FingerCode for fingerprint recognition can be
(a) (b) (c)
Figure.2: (a), (b), (c) the detected reference point in the
original image 101_7.tif (300×300) and its 1-level and 2-
level decomposed sub-images LL1 and LL2 (Db4
wavelets used).
The fingerprint image database used in this work is
the database of FVC2004 (
/fvc2004). Four distinct databases, provided by the
organizers, constitute four benchmarks: DB1, DB2,
DB3 and DB4. Each database contains 880
fingerprints for 110 fingers, each with 8 samples.
The image format is the TIF with 256 grey levels.
The images are uncompressed with a resolution of
about 500dpi. The image size varies depending on
the database. The orientation of fingerprint is
approximately in the range [-30°, +30°] with respect
to the vertical orientation.
Each fingerprint in the database is matched with
all the other fingerprints in the 4 different databases.
A matching is labelled correct if the matched pairs
are determined as identical fingers. The recognition
rate of FingerCode used on wavelet sub-images LL1
and LL2 are shown as Table 1. The recognition rate
in LL1 is higher than in LL2 over all the databases.
It can achieve 96.3% in the wavelet domain LL1
when we used the database DB1.
To compare the performance of the proposed
method with a typical method, a receiver operating
characteristic (ROC) is used. ROC is a plot of
Genuine Acceptance Rate (GAR) against False
Acceptance Rate (FAR). Fig.3. compares the ROCs
of the method based on FingerCode in ref (Sha, et
al., 2003) with the proposed algorithm on database
DB1 and DB2. Since the ROC curve of our
proposed algorithm is above the curve of method
(Sha, et al., 2003), we consider our algorithm
performs better than the method (Sha, et al., 2003)
on these databases. For example, at a 0.5% FAR, the
GAR of our proposed algorithm is 95.1%, while
minutiae-based 90.3% on database DB1.
Table1: Comparison of the recognition rate in the wavelet
domain LL1 and LL2 to different database DB1-4.
Fig.3. The ROC curve comparing the performance of the
proposed methods with method (Sha, et al., 2003) based
on (a) DB1, (b) DB2.
In this work, a novel method for fingerprint
recognition using Fingercodes in the wavelet
transform domain is proposed. One marked
advantages of our proposed method is many
conventional pre-processing such as smoothing,
binarization, thinning and restoration are not
necessary. Also, since the wavelet sub-band image
size for processing is reduced comparing to the
original image, the computational complexity is also
lowered. Above all, experimental results show that
the proposed method is outperforming the typical
Fingercode method in terms of accuracy and
In addition, since the work is based on the
sophisticated FingerCode technology and the
reliable reference point detection algorithm in ref
(Sha, et al., 2003), the proposed recognition
algorithm can be robust to noise.
This work is supported by Research Centre for
Advanced Image and Information Technology at
Chonbuk National University, Korea.
Jang X. and Yau W.Y., 2000. in Proc. Int. Conf. on
Pattren Recognition(15th), Vol.2,pp. 1024-1045.
Liu J., Huang Z., and Chan K., 2000. in Proc. Int. Conf.
on Image Processing, Vol. 2, pp 427-430.
Jain, A.K., Prabhakar, S. Lin H., 1999. IEEE
Transactions on Pattern Analysis and Machine
Intelligence, Vol. 21, Issue 4, April, pp348 – 359.
Jain A.K., Prabhakar S., Lin H., Pankanti, S., 1999. IEEE
Conf. on Computer Vision and Pattern Recognition,
Vol. 2, pp23-25.
Jain A.K., Prabhakar S., Hong L., Pankanti S., 2000.
IEEE Transactions on Image Processing, vol9, no.5,
pp 846-859.
Sha L.F., Zhao F, Tang X.O., 2003. International
Conference on Image Processing, Vol. 2, pp:II-895-8,
Selvaraj H., Arivazhagan S., Ganesan L., 2003. Fifth
International Conference on Computational
Intelligence and Multimedia Applications, 27-30
pp430 – 435.
DB1 96.3% 95.1%
DB2 95.2% 94.8%
DB3 94.7% 91.7%
DB4 95.4% 93.6%
Tico M., kuosmanen P., and Saarien J., 2001. Electronics
Letters, vol.37, no.1, pp.21-22.
Tico M., Immonen E., Ramo P., Kuosmanen P., Saarinen
J., 2001. IEEE International Symposium on Circuits
and Systems, Vol. 2, pp 21 – 24.
Fung Y.H., Chan Y.H., 2004. “Fingerprint recognition
with improved wavelet domain features,” Proceedings
of Intelligent Multimedia, Video and Speech
Processing, pp: 33 – 36.
Lee W.K., chung J.H., 1997. IEEE International
Symposium on Circuits and Systems, vol.2, pp 1201 -
Mokju M., Abu-Bakar S.A.R., 2004. International
Conference on Computer Graphics, Imaging and
Visualization, pp149 – 152.
Mallat S., 1998, the book, Academic Press.
Maltoni D., et al., 2003, the book,Springer.