Secure fingerprint identification based on structural and microangiographic optical coherence tomography
Optical coherence tomography (OCT) allows noncontact acquisition of fingerprints and hence is a highly promising technology in the field of biometrics. OCT can be used to acquire both structural and microangiographic images of fingerprints. Microangiographic OCT derives its contrast from the blood flow in the vasculature of viable skin tissue, and microangiographic fingerprint imaging is inherently immune to fake fingerprint attack. Therefore, dual-modality (structural and microangiographic) OCT imaging of fingerprints will enable more secure acquisition of biometric data, which has not been investigated before. Our study on fingerprint identification based on structural and microangiographic OCT imaging is, we believe, highly innovative. In this study, we performed OCT imaging study for fingerprint acquisition, and demonstrated the capability of dual-modality OCT imaging for the identification of fake fingerprints.
Secure fingerprint identification based on
structural and microangiographic optical
XUAN LIU,1,*FARZANA ZAKI,1YAHUI WANG,1QIONGDAN HUANG,1XIN MEI,1AND JIANGJUN WANG1,2
1Department of Electrical and Computer Engineering, New Jersey Institute of Technology, Newark, New Jersey 07102, USA
2School of Mechanical Engineering, Shandong University of Technology, Zibo 255049, China
*Corresponding author: firstname.lastname@example.org
Received 15 December 2016; revised 1 February 2017; accepted 15 February 2017; posted 16 February 2017 (Doc. ID 282921);
published 8 March 2017
Optical coherence tomography (OCT) allows noncontact acquisition of fingerprints and hence is a highly prom-
ising technology in the field of biometrics. OCT can be used to acquire both structural and microangiographic
images of fingerprints. Microangiographic OCT derives its contrast from the blood flow in the vasculature of
viable skin tissue, and microangiographic fingerprint imaging is inherently immune to fake fingerprint attack.
Therefore, dual-modality (structural and microangiographic) OCT imaging of fingerprints will enable more se-
cure acquisition of biometric data, which has not been investigated before. Our study on fingerprint identification
based on structural and microangiographic OCT imaging is, we believe, highly innovative. In this study, we
performed OCT imaging study for fingerprint acquisition, and demonstrated the capability of dual-modality
OCT imaging for the identification of fake fingerprints. © 2017 Optical Society of America
OCIS codes: (110.4500) Optical coherence tomography; (110.6880) Three-dimensional image acquisition.
Fingerprint identification has been a major biometrics technol-
ogy for over a century [1,2]. It has been widely adopted by
government agencies all over the world. The database of finger-
prints is the most comprehensive biometric database for per-
sonnel identification, and hence fingerprint identification
will continue to play a crucial role in the global market of bio-
metrics. However, current fingerprint acquisition technology
lacks robustness when the surface of the fingertip is not in
its optimal condition. Wet, dry, or damaged fingers may result
in suboptimal image quality. Moreover, current fingerprint
acquisition technology may be vulnerable to spoof attacks and
Optical coherence tomography (OCT), a three-dimensional
cross-sectional imaging modality with microscopic resolution,
has been used for fingerprint acquisition [3–6]. OCT has the
potential to allow noncontact acquisition of fingerprints, and
therefore is resistant to personal identity theft. Moreover,
OCT captures the internal fingerprint pattern, and high-quality
fingerprint images can be obtained under different surface con-
ditions. In addition, OCT, without any hardware modification,
can generate a fingerprint image by detecting subsurface vascu-
lature that follows the surface fingerprint [7,8]. Fingerprint im-
ages derived from microangiographic OCT signals can only be
acquired from viable skin tissue, and hence are immune to fake
Previously, OCT has been investigated for fingerprint ac-
quisition as an alternative to commercial fingerprint scanners.
Aum et al. performed volumetric structural imaging for live ac-
quisition of internal fingerprints with automated detection of
subsurface layers . Liu and Chen performed Doppler OCT
imaging to capture the fingertip vascular . However, the
combination of structural and vasculature imaging for finger-
print identification has not been investigated. The novelty of
this study is the acquisition of microangiographic OCT images
to confirm the fingerprint is obtained from viable tissue in ad-
dition to structural imaging for fingerprint identification.
In this study, we performed simultaneous dual-modality (struc-
tural and microangiographic) fingerprint imaging using volumetric
OCT data, which has not been demonstrated before, to the best
of our knowledge. In addition to structural imaging for fingerprint
identification, we acquired microangiographic OCT images to
confirm the fingerprint was obtained from viable tissue.
2. OCT SYSTEM FOR FINGERPRINT IMAGING
A. OCT Engine
In this study, we used a Fourier domain OCT (FD OCT)
system in which the interference signal was detected by a
Research Article Vol. 56, No. 8 / March 10 2017 / Applied Optics 2255
1559-128X/17/082255-05 Journal © 2017 Optical Society of America
spectrometer. Details about this system can be found in our pre-
vious publication . The configuration of our OCT system for
fingerprint acquisition is illustrated in Fig. 1. Briefly, the system
uses a near-infrared broadband light source (superluminescent
diode, 1.3 μm central wavelength and 100 nm bandwidth).
The superluminescent diode illuminates a fiber-optic Michelson
interferometer with a sample arm and a reference arm. A lens
(Thorlabs, AC254-100-C, 100 mm focal length, achromatic
doublet) is used in the sample arm to focus the probing beam
and collect photons backscattered from the fingertip. Light re-
turned from the sample and the reference mirror interferes
and is detected by a CMOS InGaAs camera (SUI1024LDH2,
Goodrich). Signals from the camera are streamed to the host
computer (Dell Precision T7600) via a frame grabber (PCIe-
1433, National Instrument). All computation tasks are per-
formed in parallel using graphic processing units (GPUs). The
theoretical axial resolution (δz) of the OCT system is determined
by the central wavelength of the light source (λ01.3 μm)
and the bandwidth of the light source (Δλ100 nm):
0∕Δλ7.4 μm. We characterized the axial point
spread function of the system using OCT data obtained from
a mirror and estimated the axial resolution to be 8 μm. The lateral
displacement between two adjacent Ascans is approximately
25 μm. This corresponds to a 1014 dots per inch (DPI) and
hence provides higher sampling density compared to many com-
mercially available HD fingerprint scanners. Moreover, the sam-
pling is above Nyquist rate, because the lateral resolution of the
current imaging setting is approximately 53 μm. The lateral field-
of-view for fingerprint imaging is approximately 10 mm by
8 mm, determined by the voltage applied to the galvanometer
for beam scanning and the known focal length of the imaging
objective. The system has a 2.5 mm imaging depth and a
92 kHz Ascan rate determined by the camera. Volumetric
OCT data is acquired in synchronization with the lateral scanning
of a light beam by a pair of galvanometers. We developed
high-speed software based onGPU(NVIDIAGeforceGTX
780, 2304 CUDA cores at 0.9 GHz, 3 GB graphics memory)
that processes more than 100,000 Ascans per second .
Therefore, the imaging speed of the OCT system is limited by
the camera speed rather than signal processing. Images in the pa-
per are presented without any compensation or enhancement.
B. Structural Fingerprint Imaging
The OCT system performs two-dimensional (2D) lateral scan-
ning on the surface of the fingertip. Multiple cross-sectional
images [Fig. 2(a)] are obtained to form a three-dimensional
(3D) data set (IOCTx; y;z). Volumetric OCT data are further
reduced to a 2D en face image for the visualization of the finger-
print, by averaging pixel values along the zdirection. However,
when signal intensities from all the pixels in an Ascan are aver-
aged, the resultant en face image has limited contrast, because
the internal fingerprint pattern exists at a limited depth range at
the papillary layer immediately under the epidermis layer, while
the OCT signal at pixels corresponding to the surface of the
skin is much larger than the signal at other spatial locations
[Fig. 2(b)]. To generate a high-quality image of an internal
fingerprint, the pixel corresponding to skin surface is identified
(Zsurface) in each Ascan by a peak search algorithm implemented
in real time using GPU. Afterwards, the average of OCT signal
within depth range between Zstart and Zstart Zwindow is ob-
tained to generate a high-contrast en face fingerprint image
[Eq. (1)]. Zstart is chosen to be Zsurface 50 μm, because
stratum corneum, the outmost layer of the skin, has a thickness
ranging from 10 to 40 μm. Zwindow is chosen to be 500 μm,
because the epidermis layer of skin at the fingertip ranges from
100 to 500 μm
IOCTx; y;zdz: (1)
To demonstrate that averaging within a selected depth range
allows better visualization of an internal fingerprint, we show an
en face fingerprint image obtained by averaging all the pixels in
Fig. 2(c), and show an en face fingerprint image obtained by
averaging the pixels from Zstart to Zstart Zwindow in Fig. 2(d).
Clearly, Fig. 2(c) shows artifacts due to surface glare and limited
C. Microangiographic Fingerprint Imaging
To obtain a microangiographic fingerprint image, we acquired
multiple cross-sectional (Bscan) OCT images sequentially
at the same elevation position: Ix; y; z; tijyy0where
Fig. 1. Configuration of the OCT system for fingerprint imaging
(FC, fiber-optic coupler).
Fig. 2. (a) Cross-sectional OCT image obtained from the fingertip of
a healthy volunteer (E, epidermis; D, dermis; SC, stratum corneum);
(b) depth-resolved signal profile obtained by averaging the cross-sectional
image along lateral direction; (EDJ, epidermis-dermis junction); (c) en
face image obtained by averaging all the pixels in individual Ascans; (d) en
face image obtained by averaging pixels within a specific depth range for
internal fingerprint acquisition. Scale bars indicate 1 mm.
2256 Vol. 56, No. 8 / March 10 2017 / Applied Optics Research Article
i1;2;3;…8. Afterwards, the variation of signal amplitude at
the same pixel in different frames is calculated [Eq. (2) where
N8][11,12]. The temporal variation map of the speckle
pattern highlights blood vessels, because an OCT signal from
blood vessels fluctuates drastically over time due to the motion
of flowing particles within the bloodstream (Fig. 3). Isv x; y; z
obtained at different elevation plane yyiforms a volumetric
data set and a 2D vascular fingerprint image is hence obtained
by averaging speckle variance OCT (svOCT) signal along the
axial direction [Eq. (3)]. The microvasculature follows the sur-
face fingerprint as demonstrated in OCT imaging as well as
scanning electron microscopy (SEM) [7,8]. In addition, a vas-
culature fingerprint image is inherently insensitive to the sur-
face topology of the skin, because vessels do not exist in the
epidermis layer. Therefore, Eq. (3) simply averages all the pixels
in an Ascan of the microangiographic OCT data
Isvx; y;z ﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ
i1Ix; y; z; ti;(2)
Isv;enfacex; yZIsvx; y;zdz: (3)
OCT images were obtained from a healthy volunteer.
Figure 4(a) shows the structural fingerprint image obtained
using Eq. (1). Figure 4(b) shows the microangiographic finger-
print image obtained using Eqs. (2) and (3). Notably, the gray-
scale image shown in Fig. 4(b) was obtained by inverting the
vasculature image to have a similar appearance as the structural
image shown in Fig. 4(a). Figures 4(a) and 4(b) were normal-
ized to their maximum values. For comparison, we also
obtained an image of the same finger using a commercial scan-
ner (SecuGen Hamster IV, 508 DPI), as shown in Fig. 4(c).
We acquired Figs. 4(a)–4(c) from the same subject. Figure 4(c),
acquired from a commercial scanner, appears to be different
from Figs. 4(a) and 4(b), because the image formation mecha-
nism was different from OCT, and the region scanned was also
slightly displaced from that of OCT imaging. To further val-
idate the consistency between fingerprint images obtained from
different modalities, we overlaid the structural OCT fingerprint
and the vasculature OCT fingerprint with the image obtained
by the commercial scanner in Figs. 5(a) and 5(b). In Fig. 5(a),
the upper half of the image in gray scale shows the structural
OCT image. The lower half of the image is based on the RGB
color model where the red, green, and blue channels are popu-
lated by a coregistered structural OCT signal, vasculature OCT
signal, and image data from the commercial scanner. Similarly,
the upper half of Fig. 5(b) shows the corresponding region of
the vasculature OCT image, while the lower half is the overlaid
RGB image. Clearly, structural and microangiographic images
obtained from OCT are consistent with the fingerprint ob-
tained from the commercial scanner.
To demonstrate the capability of OCT in differentiating real
fingerprints of viable skin from fake fingerprints, we performed
an imaging study on a real fingerprint and a fake fingerprint. By
performing OCT scanning on the fingertip of a healthy volun-
teer, we acquired both a structural OCT image [Fig. 6(a)] and a
vasculature OCT image [Fig. 6(b)]. Notably, in Fig. 6(b), the
bottom part of Fig. 6(a) (the structural OCT image of a real
fingerprint) is stitched to the microangiographic image to dem-
onstrate the consistency. Afterwards, we applied a small amount
of pencil lead to the surface of the same finger and compressed
the finger against a glass window to generate a fake fingerprint.
The approach we took to generate the fake fingerprint was sim-
ple and fast. Therefore, the fake fingerprint had precisely the
same morphological characteristics as the true finger under
OCT examination, and could be used to validate our technol-
ogy in vasculature fingerprint detection. The structural OCT
image obtained from the fake fingerprint is shown in Fig. 6(c).
Similar to Fig. 6(b), the bottom part of Fig. 6(a) (the structural
OCT image of the real fingerprint) is stitched to the structural
image obtained from the fake fingerprint. For the fake finger-
print, we also obtained the speckle variance for microangio-
graphic imaging using the real-time GPU software, but the
result [Fig. 6(d)] does not show any recognizable pattern.
This is because the fake fingerprint lacks real vasculature struc-
ture to provide image contrast in the microangiographic image.
Fig. 3. Cross-sectional speckle variance of OCT image highlighting
blood vessels (red arrows). The scale bars indicate 0.5 mm.
Fig. 4. Fingerprint images. (a) Structural OCT image; (b) vascula-
ture image; (c) fingerprint obtained from the commercial scanner.
Fig. 5. (a) Structural OCT image (upper) overlaid with vasculature
OCT image and the image obtained from the commercial scanner
(lower); (b) vasculature OCT image (upper) overlaid with structural
OCT image and the image obtained from the commercial scanner
Research Article Vol. 56, No. 8 / March 10 2017 / Applied Optics 2257
Results in Fig. 6clearly demonstrate that the capability of
vasculature imaging of OCT allows more secure fingerprint ac-
quisition. This was further validated by correlating fingerprints
obtained from different imaging modalities using commercial
software (SecuGen FDx SDK Pro for Windows) to match dif-
ferent fingerprints. We selected the structural OCT image of
the real fingerprint as the reference image. The match of the
structural OCT images between real and fake fingerprints
was established by the software, as shown in the graphic inter-
face of the software [Fig. 7(a)]. The result in Fig. 7(a) indicates
that the risk of a fake fingerprint attack exists in any fingerprint
scanning technology. We further used the software to match
the structural OCT image of the real fingerprint with the
microangiographic images obtained from the real and fake
fingerprint, as shown in Figs. 7(b) and 7(c), respectively.
Figure 7(b) shows that the microangiographic image of the real
fingerprint matches the structural OCT image of the real
fingerprint. In comparison, Fig. 7(c) shows that the microan-
giographic image of the fake fingerprint does not match the
structural OCT image of the real fingerprint.
4. CONCLUSION AND DISCUSSION
In this study, we performed 3D OCT imaging to acquire
structural and microangiographic fingerprints, which has not
been demonstrated before and is highly innovative. Image
reconstruction was achieved in real time using GPU. Our re-
sults showed that both structural and microangiographic finger-
prints acquired by OCT were consistent with the image
obtained from the commercial scanner. We further demon-
strated that microangiographic OCT imaging could differenti-
ate a real fingerprint in viable skin tissue from a fake fingerprint
without vasculature. Therefore, dual-modality OCT imaging
of structural and microangiographic fingerprints allow more
secure acquisition of biometrics data. Despite the advantage
of the OCT imaging for fingerprint identification, the OCT
system is more expensive than a fingerprint scanner, which
may limit its use for routine personnel identification.
In this study, we acquired OCT fingerprint data from a vol-
unteer to demonstrate the feasibility of OCT for both structural
and microangiographic imaging of a fingerprint. To validate the
accuracy of OCT in fingerprint recognition, we will have to
establish a larger image bank by performing OCT scanning
on more human subjects in our future study.
Funding. China Scholarship Council (CSC); Natural
Science Foundation of Shandong Province (ZR2014JL027);
New Jersey Institute of Technology.
1. S. Prabhakar, S. Pankanti, and A. K. Jain, “Biometric recognition:
security and privacy concerns,”IEEE Security Privacy 1,33–42
2. D. Maltoni, D. Maio, A. K. Jain, and S. Prabhakar, Handbook of
Fingerprint Recognition (Springer, 2003).
3. D. Huang, E. A. Swanson, C. P. Lin, J. S. Schuman, W. G. Stinson, W.
Chang, M. R. Hee, T. Flotte, K. Gregory, C. A. Puliafito, and J. G.
Fujimoto, “Optical coherence tomography,”Science 254, 1178–1181
4. Y. Cheng and K. V. Larin, “Artificial fingerprint recognition by using
optical coherence tomography with autocorrelation analysis,”Appl.
Opt. 45, 9238–9245 (2006).
5. M. Liu and T. Buma, “Biometric mapping of fingertip eccrine glands
with optical coherence tomography,”IEEE Photon. Technol. Lett.
22, 1677–1679 (2010).
6. J. Aum, J. Kim, and J. Jeong, “Live acquisition of internal fingerprint
with automated detection of subsurface layers using OCT,”IEEE
Photon. Technol. Lett. 28, 163–166 (2016).
7. S. Sangiorgi, A. Manelli, T. Congiu, A. Bini, G. Pilato, M. Reguzzoni,
and M. Raspanti, “Microvascularization of the human digit as studied
by corrosion casting,”J. Anat. 204, 123–131 (2004).
8. G. Liu and Z. Chen, “Capturing the vital vascular fingerprint with opti-
cal coherence tomography,”Appl. Opt. 52, 5473–5477 (2013).
Fig. 6. Fingerprint images. (a) Structural OCT image from real fin-
ger; (b) vasculature OCT image from real finger; (c) structural OCT
image from fake fingerprint; (d) svOCT image from fake fingerprint.
Fig. 7. (a) Structural OCT image of fake fingerprint matches that of
a real finger; (b) vasculature OCT image of real fingerprint matches
structural OCT image of a real finger; (c) vasculature OCT image of
fake fingerprint does not match structural OCT image of a real finger.
2258 Vol. 56, No. 8 / March 10 2017 / Applied Optics Research Article
9. X. Liu, M. Kirby, and F. Zhao, “Motion analysis and removal in intensity
variation based OCT angiography,”Biomed. Opt. Express 5,
10. F. Zaki, I. Hou, D. Cooper, D. Patel, Y. Yang, and X. Liu, “High-
definition optical coherence tomography imaging for noninvasive
examination of heritage works,”Appl. Opt. 55, 10313–10317 (2016).
11. A. Mariampillai, M. K. K. Leung, M. Jarvi, B. A. Standish, K. Lee, B. C.
Wilson, A. Vitkin, and V. X. D. Yang, “Optimized speckle variance
OCT imaging of microvasculature,”Opt. Lett. 35, 1257–1259 (2010).
12. X. Liu, K. Zhang, Y. Huang, and J. U. Kang, “Spectroscopic-speckle
variance OCT for microvasculature detection and analysis,”Biomed.
Opt. Express 2, 2995–3009 (2011).
Research Article Vol. 56, No. 8 / March 10 2017 / Applied Optics 2259