Professional Documents
Culture Documents
National Laboratory of Pattern Recognition, Institute of Automation, Chinese Academy of Sciences, P.O. Box 2728,
Beijing 100080, PR China
Received 17 June 2002; received in revised form 8 January 2003
Abstract
Fingerprint image enhancement is an essential preprocessing step in ngerprint recognition applications. In this
paper, we propose a novel lter design method for ngerprint image enhancement, primarily inspired from the traditional Gabor lter (TGF). The previous ngerprint image enhancement methods based on TGF banks have some
drawbacks in their image-dependent parameter selection strategy, which leads to artifacts in some cases. To address this
issue, we develop an improved version of the TGF, called the modied Gabor lter (MGF). Its parameter selection
scheme is image-independent. The remarkable advantages of our MGF over the TGF consist in preserving ngerprint
image structure and achieving image enhancement consistency. Experimental results indicate that the proposed MGF
enhancement algorithm can reduce the FRR of a ngerprint matcher by approximately 2% at a FAR of 0.01%.
2003 Published by Elsevier Science B.V.
Keywords: Fingerprints; Enhancement; Traditional Gabor lter; Modied Gabor lter; Parameter selection; Low pass lter; Band pass
lter
1. Introduction
Fingerprint recognition is being widely applied
in the personal identication for the purpose of
high degree of security. However, some ngerprint
images captured in variant applications are poor in
quality, which corrupts the accuracy of ngerprint
recognition. Consequently, ngerprint image enhancement is usually the rst step in most automatic ngerprint identication systems (AFISs).
There have existed a variety of research activities along the stream of reducing noises and increasing the contrast between ridges and valleys
in the gray-scale ngerprint images. Some approaches are implemented in spatial domain, others
in frequency domain. OGorman and Nickerson
(1989) and Mehtre (1993) performed ngerprint
image enhancement based on directional lters;
Maio and Maltoni (1998) employed neural
network in minutiae ltering; Almansa and
Lindeberg (2000) enhanced them in scale space;
ARTICLE IN PRESS
2
Fig. 1. A ngerprint image and corresponding ridge and valley topography. The top-right region can be approximately treated as a
sinusoidal plane wave, but never the bottom-left.
ARTICLE IN PRESS
J. Yang et al. / Pattern Recognition Letters xxx (2003) xxxxxx
"
#!
1 x2/ y/2
2px/
cos
2 r2x r2y
T
2
x/ x cos / y sin /
y/ x sin / y cos /
ARTICLE IN PRESS
4
Fig. 2. The TGF and its response represented in spatial and frequency domain.
x
T1 =2 T2 =2
T1 =2 T2 =2
6
ARTICLE IN PRESS
J. Yang et al. / Pattern Recognition Letters xxx (2003) xxxxxx
f x
8
>
06 x6T1 =4
cos 2px
>
T1
>
>
>
<
2 =4
T1 =4 < x < T1 =4 T2 =2
cos 2pxT1T=4T
2
>
>
>
>
>
: cos 2pxT1 =2T2 =2
T1 =4 T2 =26 x6T1 =2 T2 =2
T1
Fig. 4. Our MGF and its response represented in spatial and frequency domain.
ARTICLE IN PRESS
6
g0 x; y; T1 ; T2 ; / h0x x; T1 ; T2 ; / h0y y; /
(
)
!
x2/
exp 2 F x/ ; T1 ; T2
2rx
(
!)
y/2
exp 2
8
2ry
The frequency representation of the MGF is no
longer a band pass lter passing only one central
frequency component, but a band pass lter associated with a bank of low pass lters (see Fig. 4).
The associated low pass lters are benecial to
passing the useful low frequency components.
Therefore, our MGF can more straightforwardly
express the texture characteristics of ngerprint
images than the TGF.
3. Parameter selection
Parameter selection plays a crucial role in the
use of the TGF and has long been a research focus
in the eld of image processing. However, the
computation of lter coecients is very complex
(Bovick et al., 1990). For texture analysis, some
principles of parameter selection are proposed (e.g.
Jain and Farrokhnia, 1991; Clausi and Jernigan,
2000) based on comparison between the output
of the human visual system and the Gabor lter
response. Responsible for the specic ngerprint image enhancement, parameter selection also
needs to be explored. In the TGF, there are ve
parameters to be specied, including the Gabor
lter orientation /, the standard deviations rx and
ry of the 2-D Gaussian function, the period T of
the assumed sinusoidal plane wave and the convolution mask size 2N 1 2N 1. Hong
et al. specied them based on the empirical data.
In our MGF, the period T is decomposed into T1
and T2 , and most of the parameters including the
convolution mask size are specied adaptively.
3.1. Orientation / of modied Gabor lter
Hong et al. (1998) rstly utilized a least mean
square estimation method to compute the orien-
PiW =2
PjW =2
uiW =2
vjW =2 2Gx u;vGy u;v
PiW =2 PjW =2
2
2
uiW =2
vjW =2 Gx u; v Gy u;v
9
Then, hi; j is regularized into the range of p=2
to p=2. Finally, the parameter / of the TGF is
chosen as the orientation of each block.
However, their block-wise scheme is coarse and
cannot obtain ne orientation eld, which tends to
corrupt the TGFs performance. In order to estimate the orientation eld more accurately, we extend their method into a pixel-wise one. For each
pixel, a block with size W W centered at the pixel
is referred to, so the orientation of each pixel can
be estimated by the formula (9). To reduce the
computational cost, a sliding window technique is
employed (Yang et al., 2002). For an image, the
orientation of the MGF is tuned to the orientation
at current pixel, and thus a low pass ltering along
the orientation and a band pass associated with
low pass ltering orthogonal to the orientation are
performed.
It needs to be emphasized that a step of
smoothing the orientation eld by a low pass lter
is necessary since sometimes the orientation eld is
distorted by noises.
3.2. Periods T1 and T2
Examining the formulas (5)(8), we draw the
conclusion that T1 =2 T2 =2 in our MGF corresponds to T in the TGF that is depicted as the
ridge distance by Hong et al. (1998) and they become the same when T1 T2 . Further investigating
the formula (8), we learn that the zero crossings of
g0 are merely determined by the oscillator F x; T1 ;
T2 . Accordingly, we specify T1 and T2 as double
ARTICLE IN PRESS
J. Yang et al. / Pattern Recognition Letters xxx (2003) xxxxxx
Fig. 5. The curve of F x; T1 ; T2 corresponding to dierent period T1 and T2 . Pixel Pa is located on a ridge, so T1 is set as the double of
ridge width 2Wr and T2 as the double of valley width 2Wv . Pixel Pb is on a valley, so T1 is set as 2Wv and T2 as 2Wr .
ARTICLE IN PRESS
8
2
exp 2rx 2 cos 2px
dx
T1
x
Q
2
2 =4
exp 2rx 2 cos 2pxT1T=4T
dx
2
R T1 =4
0
R T1 =4T2 =2
T1 =4
10
ARTICLE IN PRESS
J. Yang et al. / Pattern Recognition Letters xxx (2003) xxxxxx
3T1 =4T2 =2
T1 =4T2 =2
x2
exp 2
2rx
2px T1 =2 T2 =2
cos
T1
Table 1
Some rx s adopted in our experiments corresponding to dierent
T1 and T2
dx
0
11
T1
T2
rx
4
4
4
6
8
10
12
14
16
4; 12
14; 18
20; 28
1.5
1.6
1.8
1.8
2.5
2.7
3.0
3.5
4.0
ARTICLE IN PRESS
10
Fig. 7. Enhancement results corresponding to the ngerprint images of Fig. 8. The rst two columns are the results using the TGF with
dierent rx ; ry . The third column is the results by our MGF.
ARTICLE IN PRESS
J. Yang et al. / Pattern Recognition Letters xxx (2003) xxxxxx
11
Table 2
Fingerprint matching performance under the enhanced images
by the TGF and MGF
Filter
FAR
FRR
TGF
MGF
0.01%
0.05%
0.1%
0.15%
1%
5.9%
3.5%
5.3%
3.1%
4.3%
2.9%
3.9%
2.9%
3.1%
2.8%
Fig. 8. Some ngerprint images in our experiments. (a) is captured from an optical equipment. (b) is f23 of NIST-4. (c) is f09 of NIST4.
ARTICLE IN PRESS
12
Table 3
Comparison of time cost of ngerprint image enhancement
(based on P4 1.3 GHz, 128 M RAM PC)
Image resolution (pixel)
TGF (N 5,
ry 4:0) rx
MGF
224 288
256 256
364 256
400 376
1.8
1.8
2.0
2.2
0.90
0.94
1.13
1.76
0.92
1.01
1.22
1.89
6. Conclusion
In this paper, a MGF has been proposed for
ngerprint image enhancement. The modication
of the TGF can make the MGF more accurate in
preserving the ngerprint image topography. And
a new scheme of adaptive parameter selection for
the MGF is discussed. This scheme leads to the
image-independent advantage in the MGF. Although there are still some intermedial parameters
determined by experience, a step of image normalization can compensate the drawback.
However, some problems need to be solved in
the future. A common problem of the MGF and
TGF is that both fail when image regions are
contaminated with heavy noises. In that case, the
orientation eld can hardly be estimated and accurate computation of ridge width and valley
width is prohibitively dicult. Therefore, a step of
segmenting these unrecoverable regions from the
original image is necessary, which has been explored in Hongs work to some extent.
Acknowledgements
The authors are highly grateful to the anonymous reviewers for their signicant and constructive critiques and suggestions, which improve the
paper very much. This work was partially supported by Hundred Talents Programs of the Chinese Academy of Sciences, the Natural Science
Foundation of China, Grant No. 60172056 and
697908001, and Watchdata Digital Company. We
acknowledge that the experiments in this research
are conducted on the ngerprint database from the
NIST, University of Bologna and FVC2000. We
References
Almansa, A., Lindeberg, T., 2000. Fingerprint enhancement by
shape adaptation of scale-space operators with automatic
scale selection. IEEE Trans. Image Process. 9 (12), 2027
2042.
Bovick, A.C., Clark, M., Geisler, W.S., 1990. Multichannel
texture analysis using localized spatial lters. IEEE Trans.
Pattern Anal. Machine Intell. 12 (1), 5573.
Clausi, D.A., Jernigan, M.E., 2000. Designing Gabor lters for
optimal texture separability. Pattern Recognition 33 (11),
18351849.
Daugman, J.G., 1980. Two-dimensional spectral analysis of
cortical receptive eld proles. Vision Research 20, 847856.
Daugman, J.G., 1985. Uncertainty relation for resolution in
space, spatial frequency, and orientation optimized by twodimensional visual cortical lters. J. Optical Soc. Amer. 2
(7), 11601169.
Gabor, D., 1946. Theory of communication. J. IEE 93, 429
457.
Greenberg, S., Aladjem, M., Kogan, D., Dimitrov, I., 2000.
Fingerprint image enhancement using ltering techniques.
In: Proc. 15th Internat. Conf. on Pattern Recognition III,
Barcelona, Spain, pp. 326329.
Haralick, R.M., 1984. Digital step edges from zero crossing of
second directional derivatives. IEEE Trans. Pattern Anal.
Machine Intell. PAMI-6 (1), 5868.
Haralick, R.M., Watson, L.T., Laey, T.J., 1983. The topographic primal sketch. Internat. J. Robotics Research 2 (1),
5072.
Hong, L., Wan, Y., Jain, A.K., 1998. Fingerprint image
enhancement: Algorithm and performance evaluation.
IEEE Trans. Pattern Anal. Machine Intell. 20 (8), 777789.
Jain, A.K., Farrokhnia, F., 1991. Unsupervised texture segmentation using Gabor lters. Pattern Recognition 24 (12),
11671186.
Jiang, X., 2001. A study of ngerprint image ltering. In: Proc.
8th Internat. Conf. on Image Processing (Biometrics),
Thessaloniki, Greece, pp. 238241.
Kamei, T., Mizoguchi, M., 1995. Image lter design for
ngerprint enhancement. In: Proc. Internat. Symp. on
Computer Vision, Coral Gables, Finland, pp. 109114.
Maio, D., Maltoni, D., 1998. Neural network based minutiae
ltering in ngerprint. In: Proc. 14th Internat. Conf. on
Pattern Recognition, Brisbane, Australia, pp. 16541658.
Mehtre, B.M., 1993. Fingerprint image analysis for automatic identication. Machine Vision, Appl. 22 (6), 124
139.
OGorman, L., Nickerson, J.V., 1989. An approach to ngerprint lter design. Pattern Recognition 22 (1), 2938.
ARTICLE IN PRESS
J. Yang et al. / Pattern Recognition Letters xxx (2003) xxxxxx
Ratha, N.K., Karu, K., Chen, S., Jain, A.K., 1996. A realtime matching system for large ngerprint database. IEEE
Trans. Pattern Anal. Machine Intell. 18 (8), 799
813.
Sherlock, B.G., Monro, D.M., Millard, K., 1994. Fingerprint enhancement by directional Fourier ltering. IEE
Proc. Vision Image Signal Process 141 (2), 8794.
Tico, M., Kuosmanen, P., 1999. A topographic method for
ngerprint segmentation. In: Proc. 6th Internat. Conf. on
Image Processing, Kobe, Japan, pp. 3640.
13