ArticlePDF Available

A fingerprint matching algorithm using bit-plane extraction method with phase-only correlation


Abstract and Figures

This paper introduces a new method in fingerprint feature extraction based on bit-plane. A bit-plane image requires smaller storage space than a greyscale image. Region of interest (ROI) of a fingerprint image is extracted by using a modified blob analysis method, and then the core point of the ROI is detected for dimension reduction process. Before bit-plane is extracted, the fingerprint image is enhanced by using Fourier transform. Bit-plane 7 of the enhanced image is used as the input for fingerprint matching with phase-only correlation (POC) function. Experiment results showed that the storage space requirement for a fingerprint database could be reduced up to 87% per image for a bit-plane image compared with a greyscale image. The proposed fingerprint matching algorithm achieves 81.16% of recognition rate on FVC2002-Db1a database and 89.78% on FingerDOS database.
Content may be subject to copyright.
44 Int. J. Biometrics, Vol. 9, No. 1, 2017
Copyright © 2017 Inderscience Enterprises Ltd.
A fingerprint matching algorithm using bit-plane
extraction method with phase-only correlation
Florence Francis-Lothai and David B.L. Bong*
Faculty of Engineering,
Universiti Malaysia Sarawak,
Kota Samarahan, Malaysia
*Corresponding author
Abstract: This paper introduces a new method in fingerprint feature extraction
based on bit-plane. A bit-plane image requires smaller storage space than a
greyscale image. Region of interest (ROI) of a fingerprint image is extracted by
using a modified blob analysis method, and then the core point of the ROI is
detected for dimension reduction process. Before bit-plane is extracted, the
fingerprint image is enhanced by using Fourier transform. Bit-plane 7 of the
enhanced image is used as the input for fingerprint matching with phase-only
correlation (POC) function. Experiment results showed that the storage space
requirement for a fingerprint database could be reduced up to 87% per image
for a bit-plane image compared with a greyscale image. The proposed
fingerprint matching algorithm achieves 81.16% of recognition rate on
FVC2002-Db1a database and 89.78% on FingerDOS database.
Keywords: bit-plane; biometric; feature extraction; fingerprint matching
algorithm; phase-only correlation; POC.
Reference to this paper should be made as follows: Francis-Lothai, F. and
Bong, D.B.L. (2017) ‘A fingerprint matching algorithm using bit-plane
extraction method with phase-only correlation’, Int. J. Biometrics, Vol. 9,
No. 1, pp.44–66.
Biographical notes: Florence Francis-Lothai received her Master of
Electronics Engineering degree from Universiti Malaysia Sarawak in 2016 and
Bachelor of Electronic and Computer Engineering degree from Universiti
Malaysia Sarawak in 2011. Her research interests include biometric
identification, image processing and computer vision.
David B.L. Bong received his PhD degree from Universiti Sains Malaysia,
MSc degree from Nanyang Technological University, Singapore and BEng
degree from Universiti Teknologi Malaysia. He is currently a Senior Lecturer
with the Faculty of Engineering, Universiti Malaysia Sarawak. His research
interests include biometric identification, image quality assessment, feature
extraction, computer vision, medical imaging and intelligent systems.
This paper is a revised and expanded version of a paper entitled ‘An analysis of
the effects of bit plane extraction in fingerprint recognition’, presented at IEEE
Conference on Systems, Process and Control (ICSPC), Kuala Lumpur,
December 12, 2014.
A fingerprint matching algorithm using bit-plane extraction method 45
1 Introduction
Recognition of human fingerprint remains a popular research field which has a significant
impact on many applications such as crime investigation, home security system, mobile
devices authentication, and so on. Over the years, the cost of fingerprint sensing devices
is decreasing and access of inexpensive computing power is increasing. These have
resulted in a massive use of fingerprint-based person recognition in government,
commercial, civilian, and financial domains (Maltoni et al., 2009a). A faster and simple
fingerprint recognition system has gained more attention from researchers due to its
advantages in many applications. The increasing number of fingerprint usage led to more
fingerprint image acquisition which requires a larger database for fingerprint storage.
Bigger memory storage is needed to store the larger sizes of fingerprint images and the
cost for such memory is also higher. Even though a greyscale fingerprint image has
smaller storage capacity than a colour image, the challenge remains to further reduce the
storage capacity of a greyscale image (Allinson, 2009). Research on the possibility of
fingerprint image file size reduction is essential to minimise the requirement of large
database storage. This has to be done with little impact on the matching accuracy so that
the performance of the recognition system does not reduce.
Bit-plane extraction is one of the popular methods used in image compression and
retrieval. Size of an image could be reduced from its greyscale to certain bit-plane and
compressed using any compression techniques (Kikuchi at al., 2009). The results are
tremendous where information of the extracted image can be retrieved almost exactly the
same as the original image regardless of the reduced size. However, each bit level of the
bit-planes carries different information which has the possibility to be widely used not
only in fingerprint matching but also in other biometric recognition such as iris, palmprint
and face.
Over the past few years, fingerprint identification algorithms are built based on
greyscale image as the input image. There are limited researches using bit-plane image in
biometric especially for fingerprint. In eight bit greyscale image there are eight bit-
planes. Each of these bit-planes carry different amount of information of the original
image. The most significant bit (MSB) carries more useful information than the least
significant bit (LSB), while the LSB contains more noise than the information (Gonzalez
and Woods, 2008). However, in different biometric applications, different levels of
bit-plane were applied. For instance, in iris recognition system the LSB and MSB were
discarded and only bit-plane 2 to 6 were used (Basit and Javed, 2007). In other
approaches, bit-plane 4 to 7 were combined in face recognition system (Ting et al., 2008),
meanwhile for palmprint recognition, bit-plane 7 outperforms other bit-plane (Lee and
Bong, 2013). Based on these facts, each of the bit-plane plays different role for different
biometric features.
Although bit-plane extraction approach shows good results in face, iris, and palmprint
(Lee and Bong, 2016) applications, it is more complicated for fingerprint matching as
fingerprint contains unique minutiae which make it difficult to be recognised. Therefore,
the use of a suitable matching method is important so that bit-plane can be useful for
fingerprint applications. Phase-only correlation (POC) function is one of the great
algorithms for fingerprint recognition (Ito et al., 2004). This method has high tendency to
be applied together with bit-plane extraction method (Francis-Lothai and Bong, 2014).
POC is a phase based method which uses the phase components in 2D Discrete
Fourier Transforms (2D DFTS) of images. This method is commonly used in computer
46 F. Francis-Lothai and D.B.L. Bong
vision applications for high accuracy image registration tasks (Kuglin and Hines, 1975).
Recently, POC method is applied in fingerprint recognition as one of the solutions for
low quality fingerprints (Zhang et al., 2006). Compared with the typical minutiae-based
fingerprint recognition method, this method is much simpler as there is no minutiae
extraction involved.
In this paper, the role of bit-plane extraction in fingerprint recognition is investigated
for the purpose of reducing the needs of large capacity for database storage. Specifically,
this research addressed two aims, first, testing on the most suitable bit-plane for
fingerprint matching. Second is to design a simple and fast fingerprint system by
combining bit-plane input with POC function. The findings are quite promising where the
recognition rate produced by the proposed fingerprint recognition system is acceptable.
Fingerprint recognition using the combination of bit-plane and POC function is also
found to be better than minutiae-based fingerprint recognition.
The remainder of this paper is organised as follows. Some of related works of the
presented algorithm are described in Section 2. The proposed fingerprint recognition
system is given in Section 3. In Section 4, performance analysis of the proposed
algorithm is presented in detail together with the results. Finally, conclusions are drawn
in Section 5.
2 Related works
There are various methods used in fingerprint recognition such as complex minutiae
vector (Tong et al., 2004), filter-bank based matching (Wen et al., 2006), convex-hulls
based matching (Wen and Guo, 2009), etc., and most of these methods have no difficulty
in matching good quality images. However, fingerprint recognition remains a challenging
problem due to the difficulty in matching low-quality and displaced fingerprints (Maltoni
et al., 2009b). A more complex algorithm is usually developed to overcome the problems.
Complexity of a fingerprint recognition system increases power consumption and
computation time for fingerprint matching, which make it inefficient.
Minutiae-based recognition is one of the famous techniques used in fingerprint
recognition. A more detailed research on fingerprint minutiae as personal identification
has emerged since the introduction of minutiae by Galton (1982). In this technique,
alignment between the template and input minutiae feature sets is searched to find the
maximum number of minutiae pairing. However, this technique itself has its own
weaknesses. One of them is, when a poor fingerprint image is compared with a better
fingerprint image, the result might not as good as comparing two sets of good quality
images. Besides, the scale of the fingerprint has to be considered too. Two sets of
fingerprint may have different scales when they are taken by different scanners operating
at different resolutions.
Jea and Govindaraju (2005) proposed the use of localised secondary features derived
from relative minutiae information. They claimed that their method could solve some of
the challenges found in fingerprint recognition such as image orientation, and spatial
distortion. In another approach, Wang et al. (2006) introduced the bank of coordinate
systems. This bank is used to realign minutiae based on the most reliable set of reference
minutiae pairs. According to Wang et al. (2006), local alignment technique which uses
only a pair of reference minutiae to align two sets of minutiae, is not only hard to be
implemented for minutiae that are far away from the reference minutiae pair, but could
A fingerprint matching algorithm using bit-plane extraction method 47
also increase the chance of fingerprint image distortion. Therefore, to increase the
accuracy of the minutiae matching, they introduced global optimum alignment approach
along with the bank of coordinate systems. Based on their results, they achieved high
matching reliability with low time cost.
Unfortunately, fingerprint matching based on minutiae is still a problem and a more
complex solution is essential (Bolle et al., 2013). An automatic fingerprint recognition
based on minutiae is most often evaluated by using similarity score. Although different
information can be exploited for noisy images, combining different similarity
contributions into a single score can be complex. Besides, there are a number of people
whose fingerprints could not be identified by the minutiae-based method. One of the
reasons is their fingerprints has special conditions, where minutiae points are hard to be
extracted by image processing. According to Ito et al. (2004), five percentage of the
human population fall into this category. A few other techniques are then introduced such
as algebraic geometry (Brass and Knauer, 2002), Hough transform (Ballard, 1981),
relaxation (Rosenfeld and Kak, 1976; Ranade and Rosenfeld, 1980), and so on. These
techniques were proposed to solve point matching problem in minutiae matching. Bishnu
et al. (2006) introduced an algorithm to perform an inexact partial point pattern matching,
however, this algorithm makes some simplifying assumptions that are not always fulfilled
by minutiae points. Relaxation approach was modified to reduce the matching
complexity; however, it slows down the process because of their iterative nature (Ton and
Jain, 1989).
In different approach, phase-based image matching is applied in fingerprint
recognition. In the past, this method was used in image registration tasks for computer
vision applications (Kuglin and Hines, 1975; Takita et al., 2004). Then, it is applied in
fingerprint matching system where it compares two fingerprint images by superimposing
the images. The global pattern of ridges and valleys of a fingerprint are the main focus in
this algorithm, where the grey-level information of the image is directly used
(Stavroulakis and Stamp, 2010). Bazen et al. (2000) explained in their research that one
of the benefits of using this technique is, it has the capability to deal with low quality
images. Moreover, it is also able to verify fingerprints that suffer from non-uniform shape
distortions. Basically, two fingerprint images are said to be the same if their correlation
peak is the highest. Lindoso et al. (2007) proposed a faster correlation-based matching
system. In their research, they applied orientation field alignment to achieve a more
accurate alignment and lesser amount of correlations. Besides, they also claimed that
their method could identify the best fingerprint image area for correlation.
Before POC function was first introduced, ordinary correlation function was used in
pattern recognition and image processing. However, ordinary correlation function tends
to produce several broad peaks and a main peak whose maximum is difficult to detect
(Cozella and Spagnolo, 2014). On the other hand, POC function yields an even sharp
maximum at the best match point; therefore it is easy to locate the distinct peak. The use
of Fourier phase information in POC on fingerprint images makes possible highly reliable
fingerprint matching for low quality fingerprints (Ito et al., 2005). This method is simpler
compared to minutiae-based recognition because minutiae extraction is not required in
the image matching measurement (Jea and Govindaraju, 2005).
Ito et al. (2004) proposed a fingerprint matching system based on POC. Their
proposed algorithm exhibits efficient identification with 1.7% of equal error rate (EER)
while EER of minutiae-based algorithm is 7%. This technique is further developed and
combined with other techniques to enhance its performance. Zhang et al. (2006) applied
48 F. Francis-Lothai and D.B.L. Bong
this technique together with Fourier-Mellin transform (FMT) in their fingerprint
matching system. The purpose of the FMT is to align two images in a shorter time. Based
on their results, they achieved EER of 3.8% and 6.2% for FVC2002 DB1 and DB3
fingerprint databases, respectively. POC function is also applied in other biometric
recognition such as palmprint recognition. Zhu et al. (2009) combined the POC method
with directional representation of palmprint. They achieved higher recognition rate and
lower EER. Besides, POC method is also applied for finger vein recognition (Mahri et al.,
2010). The EER gained in the report is quite low, i.e., 0.9803%.
Ting et al. (2008) are among the first researchers applied bit-plane extraction in
biometric applications. In their research, they use bit-plane extraction in face recognition
to reduce the complexity of computer computation as well as to extract multiple feature
data from a single image. Figure 1 illustrates the bit-plane extraction in face recognition
as proposed by Maltoni et al. (2009a). According to Ting et al. (2008), bit-plane 4 to
bit-plane 7 contain significant data, therefore, these bit-planes have a better recognition
rate than the lower order bit-planes. Their work is further expanded by analysing the
advantages of using bit-plane image over grey-level image in face recognition. In their
experiments, they achieved lower mean false acceptance rate (FAR), false rejection rate
(FRR), and half total error rate (HTER) in bit-plane image compared with grey-level
image for both AMP Face Expression Database and Yale Face Database (Ting et al.,
Figure 1 Bit-plane extraction in face recognition
Source: Ting et al. (2008)
Bit-plane extraction is also applied in other biometric recognition, such as palmprint (Lee
and Bong, 2013) and iris recognition (Basit and Javed, 2007). Recognition rates of over
80% were obtained from the palmprint recognition system proposed by Lee and Bong
(2013). Similar to face recognition, the MSB in palmprint image contributes highest
recognition rate compared with the LSB. However, it is different for iris recognition as
A fingerprint matching algorithm using bit-plane extraction method 49
observed by Basit and Javed (2007). Bit-plane 0 and 7 as shown in Figure 2 were not
used in their experiment because they claimed that these bit-planes do not contain the
relative distinct information. A normalised iris image is composed of middle frequency
component, as opposed with the idea proposed by Gonzalez and Woods (2008) where the
majority of the image information lies on the MSB (Basit and Javed, 2007).
Bonney et al. (2004) applied bit-plane extraction in their proposed iris localisation for
recognition to extract the pupillary boundary. Instead of using the most significant
bit-plane, the LSB was utilised in the application. According to Bonney et al. (2004),
bit-plane 0 provides a relatively homogenous region that is easily identifiable as the
pupil. Binary morphology was applied to the extracted bit-plane and further processed by
standard deviation windows to localise the iris. From their initial test results, it is shown
that this method has potential to be adapted for iris identification. Figure 3 shows the
bit-plane extraction in iris recognition as explained before. From the figure, it can be seen
that the pupil is successfully extracted from bit-plane 0.
Figure 2 Bit-plane extraction in normalised iris image, 1st row representing bit-plane 1, 2nd row
representing bit-plane 2 and so on, last row is the normalised irises
Source: Basit and Javed (2007)
Figure 3 Bit-plane extraction in iris recognition
Source: Bonney et al. (2004)
50 F. Francis-Lothai and D.B.L. Bong
One of the benefits of bit-plane extraction is for image compression, where a digital
image is represented by a reduced amount of data, or a reduced file size. The
compression method is based on the concept of decomposing a multilevel image into a
series of binary images. Kikuchi et al. (2009) proposed a simple bit-plane coding for
lossless image compression. In their study, it is shown that the compression method is
competitive with JPEG-LS and JPEG 2000 where the bit rates for their proposed
compression method is lower than other existing methods. The compression method is
further improved by Chen et al. (2010) where they applied context-tree modelling on the
bit-plane coding algorithm. Based on their experiment, their proposed method
outperforms other competitive methods in image compression. Both papers show that bit-
plane extraction is useful for image compression where the image is compressed without
losing too much information. Bit-plane extraction is also applied in other applications,
such as image watermarking. The image is watermarked by manipulating bit-planes of
the image. The manipulation involves LSB substitution of the first bit-plane of pixels
(Memon et al., 2014).
Based on the literature related to bit-plane, each bit-plane of biometric images has
different functionalities for recognition. For face and palmprint, bit-plane 4 to 7 are more
suitable for the feature extraction, while for iris it depends on the purpose of the
algorithms. As for fingerprint, there is still lack of applications of bit-plane as most of the
current researches focus more on minutiae-based recognition. Another reason behind the
application of bit-plane extraction in biometric recognition is the nature of a single bit-
plane image which is taking lesser storage capacity. The number of fingerprint databases
increase each year, therefore a bigger storage is needed, which means higher cost.
3 Fingerprint matching algorithm
This section describes fingerprint recognition algorithm using bit-plane extraction and
POC function. The proposed matching algorithm is summarised as follows.
Step 1 extract the ROI of s(m, n) and t(m, n) using blob analysis
Step 2 rotate s(m, n) over the angular rangeθmax θ θmax with an angle spacing 5°
Step 3 locate the core point of sθ(m, n) and tθ(m, n)
Step 4 crop sθ(m, n) and tθ(m, n) based on the located core point
Step 5 enhance cropped sθ(m, n) and tθ(m, n) using Fourier transform
Step 6 extract bit-plane, kth, of enhanced sθ(m, n) and tθ(m, n)
Step 7 calculate the POC function ˆ(,)rmn
Step 8 compute the matching score.
3.1 Fingerprint database
There are two different fingerprint databases used in the research, i.e., FingerDOS
(Francis-Lothai and Bong, 2015) and FVC2002-Db1a (Maio et al., 2002). The first
database is a self-collected database and the latter is obtained from a benchmark database.
A fingerprint matching algorithm using bit-plane extraction method 51
FingerDOS is collected to provide more fingerprint samples per finger for analysis.
Besides, this database also aims to provide samples for performance comparison of the
proposed fingerprint recognition system. FingerDOS is acquired using an optical
fingerprint sensor, i.e., SecuGen iD-USB SC, and contains 3,600 fingerprint images. The
specifications of the sensor are shown in Table 1. The 60 subjects involved in the
fingerprint acquisition are from multiple ethnics with average age of 22 years old. 56% of
the subjects are males and 44% are females. The database images are acquired from six
fingers, i.e., thumb, index and middle finger for both right and left hands of the subjects.
A total of ten samples of fingerprint image are captured from each finger. The fingerprint
images are then saved in 256 grey-level bitmap image file (BMP) with the size of 77.2kB
per image. In this research, only 600 fingerprint images are used in the analysis, i.e.,
thumb of the subjects’ right hand. Figure 4 shows samples of a thumbprint from
FingerDOS with ten different impressions of the same finger.
Table 1 Sensor specifications (U.S.B. Sc and P.S.C. Datasheet)
Specification SecuGen iD-USB SC
Image resolution 500 ppi
Image size 260×300 pixels
Platen size 16.1mm×18.2mm
Effective sensing area 13.2mm×15.2mm
Image greyscale 256 levels (8 bit)
Light source/typical lifetime Red LED/60,000 hours
Fingerprint capture time 0.2~0.5 second with Smart Capture
Figure 4 Fingerprint image from FingerDOS
FVC2002-DB1a database is selected for the analysis in this research because the database
is collected by using optical sensor, the same type as in FingerDOS. However, the sensor
is manufactured by different manufacturer, i.e., TouchView II by Identix. In this
database, there are 30 different subjects. Two fingers, which are index and middle fingers
from both hands, with eight impressions from each finger, were acquired. Total number
of images in this database is 800 images and all of the images are used in the analysis.
The size of the image from this database is 388×374 pixels and saved in TIFF format at
52 F. Francis-Lothai and D.B.L. Bong
142kB per image. Figure 5 shows image samples from FVC2002-DB1a. It can be seen
that the location of the captured images is inconsistent in term of its placement on the
sensor plate. Some of the images only have half of the fingerprint compared with the
images in FingerDOS.
Figure 5 Fingerprint image from FVC2002-DB1A
3.2 Fingerprint ROI detection by using blob analysis
The region of interest (ROI) of the fingerprint is detected and separated from its
background by using blob analysis. The background corresponds to the unwanted regions
outside the fingerprint area, which contain noise or other types of invalid fingerprint
information. This technique is modified by adding some other techniques so that it is
suitable for fingerprint ROI detection. The steps involved in this process are image
binarisation, block direction estimation, and fingerprint ROI detection.
The grayscale image of the fingerprint is binarised to transform the image to a one bit
image with binary value of zero for ridges and one for valleys. Adaptive local
thresholding method is applied to the image to perform image binarisation. Intensities of
grey levels of the fingerprint image is analysed within a local window (16×16 pixels)
over the image, so that local thresholds can be determined. The mean pixel value within
the local window is examined and any grey level above the mean value is converted to
one and all other pixels to zero.
Next, block direction for each block of the binary image is estimated to extract the
optimal dominant ridge orientation. This method is based on computation of gradients in
the fingerprint image. Following are the steps involved for each of the block.
1 Let Gx(i, j) and Gy(i, j) be the gradient magnitude in x and y direction at pixel (i, j),
respectively. The gradient magnitude is computed by using Sobel masks.
2 The least square approximation of the block direction is obtained using the following
16 16
16 16 2 2
2(, ) (, )
Gij Gij
ΣΣ −
A fingerprint matching algorithm using bit-plane extraction method 53
where Gx 0 and Gy 0. Blocks without significant information on ridges and
valleys are discarded based on the following equation:
16 16 16 16 2 2
11 11
16 16 2 2
2 (,) (,) (,) (,)
(, ) (, )
xy x y
ij ij
GijGij Gij Gij
BWW Gij Gij
== ==
ΣΣ +ΣΣ −
=×ΣΣ + (2)
The block is regarded as background block if the certainty level Bd is below a threshold.
Figure 6(a) and Figure 6(b) illustrate the binarised and orientation field of the fingerprint.
The orientation field denotes the ridges of a fingerprint. The purpose of this step is to
allow the fingerprint area to be effectively separated from the unwanted background in
the modified blob analysis.
The variance of grey levels in a direction orthogonal is computed to the orientation
field of each block. Next, holes in the binary image are filled by a flood-fill operation.
This operation changes connected background pixels to foreground pixels, and stops
when it reaches object boundaries. The corresponding output image is shown in Figure
6(c). Any unnecessary pixels in the image are removed using branchpoints operation.
This operation eliminates any branches in the image. After that, blob labelling is done
based on 8-connectivity and each area of the blob is calculated as shown in Figure 6(d).
Only the ROI is extracted from the image and others are eliminated, leaving the largest
blob as illustrated in Figure 6(e). Boundary of the fingerprint ROI is then smoothen using
convex hull algorithm, and this image is used as a mask on the original image as shown
in Figure 6(f). Figure 7(a) and Figure 7(b) show the original and post ROI extraction
image, respectively.
Figure 6 Fingerprint ROI segmentation, (a) binary image (b) orientation field (c) filled image
(d) labelled blob (e) selected blob (f) blob edge smoothen
54 F. Francis-Lothai and D.B.L. Bong
Figure 7 Comparison between the original and post ROI extraction image, (a) original image
(b) ROI image
(a) (b)
3.3 Fingerprint image alignment
Image alignment is done to improve the efficiency of the fingerprint image recognition as
fingerprint image with displacement would highly affect the performance of a fingerprint
matching system. There are three main steps involved in this process, i.e., rotation, core
point detection, and cropping.
In image rotation, the extracted ROI fingerprint image is rotated into 21 different
angle (–10° θ 10°) with angle spacing of 1° for FingerDOS and 12 different angle
(30° θ 30°) with 5° spacing for FVC2002-Db1a. Both rotations are using bilinear
interpolation method. Then, core point of the rotated fingerprint image is detected using
Poincaré index. This algorithm is based on orientation field estimation proposed by
Maltoni et al. (2009a). The core is expected to occur at the position at which region with
a largest value in centre, which also continuously connects with a largest neighbourhood
area before the value reduces to a threshold.
The core point of the fingerprint image is used as a reference point for image
cropping. The size of the cropping area is determined experimentally, which are 200×200
and 180×180 pixels, for both databases, i.e., FingerDOS and FVC2002-Db1a. To crop
the fingerprint ROI, the top-left corner coordinates of the desired ROI rectangle is
determined by using the following equations:
min 2
=− (3)
min 2
=− (4)
where xmin and ymin is the coordinate, xc and yc is the fingerprint ROI core point. Parameter
width and length are determined constants. The image is then cropped based on the three
parameters, which are top-left corner coordinates, width, and length of the image. Figure
8 illustrates the cropping process, where I is the rotated image with M × N size, and Ic is
the cropped fingerprint.
A fingerprint matching algorithm using bit-plane extraction method 55
Figure 8 ROI fingerprint image cropping
3.4 Image enhancement by Fourier transform
The cropped fingerprint image is enhanced by using Fourier transform technique. To
perform this step, the image is divided into image block of 32×32 pixels. The Fourier
transform is defined by:
11 2
(, ) (, )
ux vy
MN jπ
Fuv f x ye
−− ⎛ ⎞
=∑∑ (5)
for u = 0, 1, 2,…,31 and v = 0, 1, 2,…,31. Each of the image block is then enhanced
separately, where the Fourier transform of the block is multiplied by its power spectrum
raised to a power of k:
(, ) (, ) (, )
where F–1(F(u, v)) is given by:
11 2
(, ) (, )
ux vy
MN jπ
fxy Fuve
−− ⎛ ⎞
=∑∑ (7)
for x = 0, 1, 2,…,31 and y = 0, 1, 2,…,31. The value of k is a determined constant, which
k = 0.2. A very high value of k causes false joining of ridges, therefore it is set to an
optimal value where the appearance of the ridges is improved without giving too much
effects to the original ridges. After enhancement, some of the spurious connections
between ridges are removed and falsely broken points on the ridges are also improved.
Figure 9 shows the fingerprint before and after applying FFT.
56 F. Francis-Lothai and D.B.L. Bong
Figure 9 Image enhancement by FFT, (a) before applying FFT (b) after applying FFT
(a) (b)
3.5 Fingerprint feature extraction
Bit-plane extraction is a technique used to break a greyscale image down into a sequence
of binary images. An M × N fingerprint binary image, I can be decomposed into its eight
bit-planes by using the following equation:
(,) (,
m n R floor I m n
where i = 0, 1, 2,…,7, I(m, n) is the pixel values of greyscale fingerprint image, Ii is the i-
th bit-plane, R is the remainder, and floor(m) is round the elements to m nearest integers
less than or equal to m.
Consider an 8-bit greyscale fingerprint image and the pixel representation at the
pointed area is illustrated in Figure 10(a). The binary formats for those values are
represented in Figure 10(b). To extract bit-planes, values from the same bit-level of
binaries are grouped together. Bit-plane 0 contains the lowest order bit of all the pixels in
the image, and bit-plane 7 contains the highest order bit as shown in Figure 10(c).
Figure 10 Bit-plane extraction of a fingerprint image, (a) greyscale fingerprint image and the
pixel representation at a point (b) binary value of each pixels (c) bit-plane
representation (see online version for colours)
(a) (b) (c)
In this paper, single bit-plane is proposed for fingerprint recognition because patterns of
the fingerprint are visibly found in bit-plane 7. The remaining bit-planes contain more
noise than useful information; therefore they are not suitable for fingerprint recognition.
Figure 11 illustrates all the extracted fingerprint bit-plane images.
A fingerprint matching algorithm using bit-plane extraction method 57
Figure 11 Bit-plane feature extraction from enhanced greyscale fingerprint
Bit-plane 7 Bit-plane 6 Bit-plane 5 Bit-plane 4
Bit-plane 3 Bit-plane 2 Bit-plane 1 Bit-plane 0
3.6 POC fingerprint matching
In the fingerprint recognition stage, POC function is used as the matching tool. Basically,
this function uses the phase spectra of images. The POC of two compared fingerprint
images, namely registered s(m, n) and input fingerprint t(m, n), are calculated by using
this function and their similarity are evaluated based on the peaks produced from the
Given two M × N images, s(m, n) and t(m, n), and their index ranges are
{,,}mMM∈− (9)
{,,}nNN∈− (10)
where M > 0 and N > 0, both are integers.
Let S(u, v) and T(u, v) denote the two-dimensional Discrete Fourier transform (DFT)
of the two images. S(u, v) and T(u, v) are given by
(, ) ( , )
(, ) s
MN jvnπ
Suv smne e
=− =−
∑∑ (11)
(, ) ( , )
(, ) T
MN jvnπ
Tuv tmne e
=− =−
∑∑ (12)
where u = –M,…,M and v = –N,…,N. As(u, v) and AT(u, v) are the amplitude components,
and θs(u, v) and θT(u, v) are the phase components. The cross phase spectrum of the two
images is given by
58 F. Francis-Lothai and D.B.L. Bong
(, ) (, )
ˆ(, )
(, ) (, )
where (, )Tuv is the complex conjugate of T(u, v), and θ(u, v) denotes the phase
difference of θs(u, v) – θT(u, v). The POC function, ˆ(,)
rmn, is the two-dimensional
inverse Discrete Fourier transform (2D IDFT) of ˆ(, )
Ruv, and is given by
11 2
ˆ(,) (,)
MN πjnv
st ST
rmn R uve e
=∑∑ (14)
When s(m, n) and t(m, n) are the same image, i.e., s(m, n) = t(m, n), the POC function will
be given by
11 2
MN πjnv
rmn e e
if m n
4 Performance evaluation of the fingerprint recognition system
The experiments are designed in order to show the performance of the proposed
algorithm on two different fingerprint databases, i.e., FingerDOS and FVC2002-Db1a,
with different degree of misalignment.
Table 2 FPR and average recognition rate of the bit-planes
Bit-plane FPR (%) Average recognition rate (%)
0 57.7 60.89
1 69.5 60.92
2 61.5 64.81
3 71.5 57.85
4 70.8 59.30
5 81.5 73.14
6 30.5 83.44
7 0.1 92.81
4.1 Bit-plane selection
The most useful bit-plane for fingerprint recognition is selected based on the highest
average recognition rate produced by each of the bit-planes. To select the bit-plane, an
experiment was conducted by using FingerDOS database and 300 fingerprint images
were selected for the analysis. This experiment constitutes the training stage to determine
the best bit plane for fingerprint recognition. A table of false positive rate (FPR) achieved
A fingerprint matching algorithm using bit-plane extraction method 59
at true positive rate (TPR) of 95.7% is then tabulated in Table 2 together with the average
recognition rate produced by each of the bit-plane fingerprint image. From these results,
bit-plane 7 exhibits the best result with the least FPR of 0.1% and the highest average
recognition rate of 92.81%. Since bit-plane 7 outperformed all other bit-planes, this bit-
plane is selected for the following evaluations in the next sections of the paper.
4.2 Optimal rotation angle of misaligned fingerprint
Two different fingerprint databases are used to test the recognition system, i.e.,
FingerDOS and FVC2002-Db1a. These two databases have different degree of
misalignment. Therefore, different rotation angle is needed to readjust the fingerprint
images. Number of possible matching pairs in FingerDOS and FVC2002-Db1a are 6,000
and 6,400 pairs, respectively. The best matching pair is defined as the highest correlation
peak of the POC function obtained from matching two fingerprint images, the closer the
correlation value to 1, the better it is.
Table 3 Number of matching pairs at different angle for FingerDOS and FVC2002-Db1A
FingerDOS FVC2002-Db1a
rotation (°)
No. of matching
No. of matching
0 1814 0 3824
1 1595 5 1032
2 963 10 504
3 591 15 288
4 353 20 104
5 221 25 192
6 147 30 456
7 113
8 69
9 52
10 82
Table 3 depicts the number of best matching pairs produced at different rotation for
FingerDOS and FVC2002-Db1a databases. It is shown that most of the bit-plane images
from both databases fall in zero rotation category, with highest total number of matching
pairs of 1,814 and 3,824, respectively. In FingerDOS however, only 82 pairs of
fingerprint matching fall in rotation of 10° compared with FVC2002-Db1a which
involves 504 pairs of fingerprint matching. These results show that fingerprint images
from different fingerprint database require different number and angle of rotations.
Despite the differences in rotation angle, the proposed fingerprint recognition system is
able to handle difficult fingerprint image with misalignment.
4.3 Effects of reducing the fingerprint image area
POC is simpler than other matching algorithms but it is easily affected by the
displacement of the fingerprint image. After image rotation, size of the fingerprint image
60 F. Francis-Lothai and D.B.L. Bong
has increased. Image cropping is essential to ensure an even image size for fingerprint
matching. Two different cropping sizes for both databases were used, i.e., 180 × 180 and
200 × 200 pixels. Effects of reducing the image dimension is analysed by computing the
recognition rate and accuracy as depicted in Table 4 and Table 5, respectively.
Table 4 Recognition rates with different cropping sizes
Database/cropping size Original image size 200×200 180×180
FingerDOS 89.78% 82.69% 74.20%
FVC2002-Db1a 12.59% 72.89% 81.16%
Table 5 Recognition accuracy at threshold 0.05
Database/cropping size Original image size 200×200 180×180
FingerDOS 0.998 0.927 0.975
FVC2002-Db1a 0.991 0.998 0.990
From Table 4, the recognition rate for FVC2002-Db1a is greatly improved from 12.59%
to 72.89% and 81.16% for cropping size of 200 × 200 and 180 × 180 pixels, respectively.
The recognition accuracy of the proposed fingerprint recognition in both databases are
closer to 1, which means it has high accuracy in all different image size dimension used
in the experiments. Based on these results, it is shown that the fingerprint recognition
system is able to reject almost all the unauthorised subjects.
4.4 FAR, FRR, ROC, and EER
The fingerprint recognition system is evaluated based on genuine and imposter matching.
For genuine matching, each fingerprint image is compared to the remaining images of the
same finger, or also known as genuine attempts. For imposter attempt, the first
impression of each finger is compared to the first image of the remaining fingers. Two
fingerprints are determined to be matched if the calculated matching score is higher than
the predetermined threshold value. The results are then validated with the ground truth of
the fingerprints.
In this research, two types of errors are defined to analyse the performance of the
fingerprint recognition system, i.e., FAR and FRR. FAR is the frequency at which invalid
inputs are incorrectly accepted and defined as following:
Number of false acceptances
FAR Number of imposter attempts
In FingerDOS the number of imposter attempts is 60C2×1×1 = 1770. For FVC2002-Db1a
the number of imposter attempts is 100C2×1×1 = 4950. The number of imposter tests can
reach up to 177,000 and 316,800 for FingerDOS and FVC2002-Db1a databases,
respectively, when all the images of a subject in the databases are used for imposter test.
However, for this research only one image is set for each imposter. The number of false
acceptances is defined as the number of an imposter attempts which are accepted by the
FRR is the frequency in which the system incorrectly rejects access to a genuine
subject and is defined as the following:
A fingerprint matching algorithm using bit-plane extraction method 61
Number of false rejections
FRR Number of genuine attempts
Number of genuine attempts in FingerDOS and FVC2002-Db1a are 10C2×60 = 2700 and
gC2×100 = 2800 attempts, respectively. The number of false rejections is defined as the
number of genuine attempts rejected by the system.
Table 6 depicts the FAR and FRR produced at threshold value of 0.05 in FingerDOS
and FVC2002-Db1a. The recognition system is able to reject all the imposters in
FingerDOS with 0% FAR. However, 11.06% of the genuine attempts are rejected. In
FVC2002-Db1a, 7.93% of FAR and 21.53% of FRR are produced. For this database, the
recognition system is better in rejecting an imposter and accepting a genuine attempt.
One of the reasons for the lower performance when tested with FVC2002-Db1a is
because most of the images in the database have higher degree of misalignment and
incomplete fingerprint image.
Table 6 FAR and FRR at threshold value of 0.05
Database No. of genuine attempts No. of imposter attempts FAR FRR
FingerDOS 2,700 1,770 0% 11.06%
FVC2002-Db1a 2,800 4,950 7.93% 21.53%
The accuracy of the fingerprint recognition is quantified by ROC curve. Figure 12
illustrates the ROC curves for FingerDOS and FVC2002-Db1a. From Figure 12, when
the genuine attempts are 92%, the FAR produced in FingerDOS and FVC2002-Db1a are
0% and 48%, respectively. Higher percentage achieved in FingerDOS explained the
difficultness of the fingerprint images in FVC2002-Db1a.
Figure 12 ROC curve with the detection rate against the far
The EER curves are plotted as in Figure 13(a) and Figure 13(b) for FingerDOS and
FVC2002-Db1a databases, respectively. Table 7 depicted the EER for both databases,
where the EER is higher for FVC2002-Db1a. The system falsely rejected 18.23%
genuine fingerprint pairs and falsely accepted 18.23% of imposter fingerprint pairs. The
high incorrect acceptance and false rejection in FVC2002-Db1a are due to some
fingerprint images with bad quality and the vulnerable fingerprint recognition algorithm.
62 F. Francis-Lothai and D.B.L. Bong
Figure 13 EER curves, (a) FingerDOS (b) FVC2002-Db1a
(a) (b)
Table 7 EER in percentage
Database EER
FingerDOS 7.62%
FVC2002-Db1a 18.23%
4.5 Comparison of the proposed fingerprint recognition method with other
In order to compare the performance of the proposed fingerprint recognition system with
other existing methods, an evaluation is conducted by using bit-plane 7 as the input
fingerprint image. Benchmark fingerprint database, i.e., FVC2002-Db1a is used in the
evaluation. Table 8 depicts the comparison of recognition rates between the proposed
fingerprint recognition with a minutiae-based fingerprint recognition method. From the
table, it is shown that the proposed fingerprint recognition method has better recognition
rate, i.e., 81.16% compared to other method, i.e., 74.17%. Besides, the processing time
needed to complete one matching is lower in the proposed fingerprint recognition system
compared to minutiae-based fingerprint recognition.
Table 8 Performance comparison between the proposed fingerprint recognition method with
other method
Fingerprint recognition method Recognition rate Processing time (s)
Bit-plane with POC function 81.16% 0.12
Minutiae-based 74.17% 0.68
The evaluation of the proposed fingerprint matching algorithm is further investigated by
comparing the experiments results with fingerprint recognition methods that use
FVC2002-DB1a database in their analyses. Table 9 shows that the proposed recognition
system achieves the best results for most of the performance indicators. Although the
EER for the proposed fingerprint recognition system is higher than the rest, however in
terms of speed and image size, this system has better performance.
A fingerprint matching algorithm using bit-plane extraction method 63
Table 9 Results comparison between the proposed fingerprint recognition method with other
Fingerprint recognition method Image size (kB) EER (%) Processing time (s)
Bit-plane with POC function 18.1 18.23 0.16
SIFT (Park et al., 2008) 142 8.44 4
RCS (Feng and Cai, 2006) 142 4.09
Embedded HMM (Guo, 2005) 142 7.1
4.6 Storage requirement for greyscale and bit-plane image
One of the main features of bit-plane extraction is to discard those particular bits which
are not responsible for the texture information of the image, which allows storing the
image with less number of bits. Storing a bit-plane image with lesser number of bits saves
the storage space and also improves the data transmission rate for the image. In this
research, only bit-plane 7 is used for the fingerprint recognition system which saves a lot
of storage space. Table 10 depicts the file size of the fingerprint images in greyscale and
bit-plane form. There are clear benefits from the bit-plane extraction, where the file size
is greatly reduced compared to the original greyscale fingerprints. Besides, there is no
vibrant visual effect degradation to the bit-plane image compared with the greyscale
Table 10 Size comparison between greyscale and bit-plane image
Original image Bit-plane 7 Original file
size (kB)
Bit-plane file
size (kB)
Total size
reduced (kB)
77.2 9.75 67.5
142 18.1 123.8
5 Conclusions
In this paper, a simple and fast fingerprint recognition system is proposed by combining
bit-plane input with POC function. Bit-plane extraction of bit-plane 7 extracts the useful
information in the fingerprint image without losing too much information. This method
reduces the space requirement in database storage. This paper has demonstrated that bit-
plane feature extraction has the potential to replace greyscale image in fingerprint
recognition, and it works well with POC function. By using POC function, the execution
64 F. Francis-Lothai and D.B.L. Bong
time to match a fingerprint is greatly improved compared to minutiae-based fingerprint
This project was funded by the Ministry of Higher Education, Malaysia under research
grant number FRGS/03(03)/771/2010(52).
Allinson, N.M. (2009) ‘Fingerprint compression’, in Li, S.Z. and Jain, A.K. (Eds.): Encyclopedia of
Biometrics, pp.446–452, Springer US, Boston, MA.
Ballard, D.H. (1981) ‘Generalizing the Hough transform to detect arbitrary shapes’, Pattern
Recognition, Vol. 13, No. 2, pp.111–122.
Basit, A. and Javed, M.Y. (2007) ‘Iris localization via intensity gradient and recognition through bit
planes’, in Proceedings of the International Conference on Machine Vision (ICMV), pp.23–28.
Bazen, A.M., Verwaaijen, G.T., Gerez, S.H., Veelenturf, L.P.J. and van der Zwaag, B.J. (2000) ‘A
correlation-based fingerprint verification system’, in Proceedings of ProRISC Workshop
Circuits Systems and Signal Processings, pp.1–8.
Bishnu, A., Das, S., Nandy, S.C. and Bhattacharya, B.B. (2006) ‘Simple algorithms for partial point
set pattern matching under rigid motion’, Pattern Recognition, Vol. 39, No. 9, pp.1662–1671.
Bolle, R.M., Connell, J., Pankanti, S., Ratha, N. K. and Senior, A.W. (2013) Guide to Biometrics,
Springer, New York.
Bonney, B., Ives, R., Etter, D. and Du, Y. (2004) ‘Iris pattern extraction using bit planes and
standard deviations’, in Conference Record of the 38th Asimolar Conference on Signals,
Systems and Computers, Vol. 1, pp.582–586.
Brass, P. and Knauer, C. (2002) ‘Testing the congruence of d-dimensional point sets’, International
Journal of Computational Geometry & Applications, Vol. 12, No. 01n02, pp.115–124.
Chen, M., Fränti, P. and Xu, M. (2010) ‘Lossless bit-plane compression of images with context tree
modeling’, in International Conference on Green Circuits and Systems (ICGCS 2010),
Cozella, L. and Spagnolo, G.S. (2014) ‘Phase-only correlation function by means of Hartley
transform’, JSM Mathematics and Statistics, Vol. 1, No. 1, pp.1–8.
Feng, J. and Cai, A. (2006) ‘Fingerprint representation and matching in ridge coordinate system’, in
18th International Conference on Pattern Recognition (ICPR’06), Vol. 4, pp.485–488.
Francis-Lothai, F. and Bong, D.B.L. (2014) ‘An analysis of the effects of bit plane extraction in
fingerprint recognition’, in 2014 IEEE Conference on Systems, Process and Control (ICSPC),
Francis-Lothai, F. and Bong, D.B.L. (2015) ‘FingerDOS: a fingerprint database based on optical
sensor’, WSEAS Transactions on Information Science and Applications, Vol. 12, pp.297–304.
Galton, F. (1982) Finger Prints, Macmillan & Co., London.
Gonzalez, R.C. and Woods, R.E. (2008) Digital Image Processing, 3rd ed., Pearson/Prentice Hall,
New Jersey.
Guo, H. (2005) ‘A hidden Markov model fingerprint matching approach,’ in International
Conference on Machine Learning and Cybernatics, Vol. 8, pp.5055–5059.
Ito, K., Morita, A., Aoki, T., Higuchi, T., Nakajima, H. and Kobayashi, K. (2005) ‘A fingerprint
recognition algorithm using phase-based image matching for low-quality fingerprints’, in
IEEE International Conference on Image Processing (ICIP 2005), Vol. 2, pp.33–36.
A fingerprint matching algorithm using bit-plane extraction method 65
Ito, K., Nakajima, H., Kobayashi, K., Aoki, T. and Higuchi, T. (2004) ‘A fingerprint matching
algorithm using phase-only correlation’, IEICE Transaction Fundamentals, Vol. E87-A,
No. 3, pp.682–691.
Jea, T.Y. and Govindaraju, V. (2005) ‘A minutia-based partial fingerprint recognition system’,
Pattern Recognition, Vol. 38, No. 10, pp.1672–1684.
Kikuchi, H., Funahashi, K. and Muramatsu, S. (2009) ‘Simple bit-plane coding for lossless image
compression and extended functionalities’, in Proceedings of the 27th Picture Coding
Symposium (PCS), pp.1–4.
Kuglin, C.D. and Hines, D.C. (1975) ‘The phase correlation image alignment method’, in
Proceedings of International Conference Cybernetics and Society, pp.163–165.
Lee, T.Z. and Bong, D.B.L. (2013) ‘Palmprint recognition based on bit-plane extraction’, in
WSEAS 12th International Conference on Applied Computer and Applied Computational
Science, pp.182–186.
Lee, T.Z. and Bong, D.B.L. (2016) ‘Analysis of bit-plane images by using principal component on
face and palmprint database’, Pertanika Journal of Science and Technology, Vol. 24, No. 1,
Lindoso, A., Entrena, L., Liu-Jimenez, J. and Millán, E.S. (2007) ‘Correlation-based fingerprint
matching with orientation field alignment’, in Proceedings of the 2007 International
Conference on Advances in Biometrics (ICB’07), pp.713–721.
Mahri, N., Azmin, S., Suandi, S. and Rosdi, B.A. (2010) ‘Finger vein recognition algorithm using
phase only correlation’, in International Workshop on Emerging Techniques and Challenges
for Hand-Based Biometrics (ETCHB), Vol. 4642, pp.1–6.
Maio, D., Maltoni, D., Cappelli, R., Wayman, J.L. and Jain, A.K. (2002) ‘FVC2002: second
fingerprint verification competition’, in Proceedings of the 16th International Conference on
Pattern Recognition, Vol. 3, pp.811–814.
Maltoni, D., Maio, D., Jain, A.K. and Prabhakar, S. (2009a) ‘Fingerprint analysis and
representations’, in Handbook of Fingerprint Recognition, 2nd ed., pp.97–166, Springer,
Maltoni, D., Maio, D., Jain, A.K. and Prabhakar, S. (2009b) ‘Fingerprint matching’, in Handbook
of Fingerprint Recognition, 2nd ed., pp.167–233, Springer, London.
Memon, N.A., Keerio, Z. and Abbasi, F. (2014) ‘Dual watermarking on CT scan medical images
for content authentication and copyright protection’, in Shaikh, F.K., Chowdhry, B.S.,
Zeadally, S., Hussain, D.M., Memon, A.A. and Uqaili, M.A. (Eds.): Communication
Technologies, Information Security and Sustainable Development, Vol. 414, pp.173–183,
Springer International Publishing, Switzerland.
Park, U., Pankanti, S. and Jain, A.K. (2008) ‘Fingerprint verification using SIFT features,’ in SPIE
Defense and Security Symposium, Vol. 6944, p.69440K.
Ranade, S. and Rosenfeld, A. (1980) ‘Point pattern matching by relaxation’, Pattern Recognition,
Vol. 12, No. 4, pp.269–275.
Rosenfeld, A. and Kak, A.C. (1976) Digital Picture Processing, Academic Press, New York.
Stavroulakis, P. and Stamp, M. (2010) Handbook of Information and Communication Security.
Springer Berlin, Heidelberg.
Takita, K., Muquit, M.A., Aoki, T. and Higuchi, T. (2004) ‘A sub-pixel correspondence search
technique for computer vision applications’, IEICE Transactions on Fundamentals of
Electronics, Communications and Computer Science, Vol. E87-A, No. 8, pp.1913–1923.
Ting, K.C., Bong, D.B.L. and Wang, Y.C. (2008) ‘Performance analysis of single and combined
bit-planes feature extraction for recognition in face expression database’, in Proceedings of the
International Conference on Computer and Communication Engineering (ICCCE),
Ting, K.C., Tan, J.Y.B., Lee, T.Z. and Bong, D.B.L. (2013) ‘Face recognition by neural network
using bit-planes extracted from an image’, Journal of Information and Computational Science,
Vol. 10, No. 16, pp.5253–5261.
66 F. Francis-Lothai and D.B.L. Bong
Ton, J. and Jain, A.K. (1989) ‘Registering landsat images by point matching’, IEEE Transactions
on Geoscience and Remote Sensing, Vol. 27, No. 5, pp.642–651.
Tong, X., Tang, X., Huang, J. and Li, X. (2004) ‘Fingerprint minutiae matching based on complex
minutiae vector’, in Proceedings of the 3rd International Conference on Machine Learning
and Cybernetics, Vol. 6, pp.3731–3735.
U.S.B. Sc and P.S.C. Datasheet (2012) SC SecuGen iD -USB SC TM Model: XSDU03PSC
Datasheet, Vol. 004, pp.2–5.
Wang, W., Li, J. and Chen, W. (2006) ‘Fingerprint minutiae matching based on coordinate system
bank and global optimum alignment’, in 18th International Conference on Pattern
Recognition (ICPR’06), Vol. 4, pp.401–404.
Wen, C. and Guo, T. (2009) ‘An efficient algorithm for fingerprint matching based on convex
hulls’, in International Conference on Computational Intelligence and Natural Computing,
(CINC), Vol. 1, pp.66–69.
Wen, M.L., Liang, Y., Pan, Q. and Zhang, H.C. (2006) ‘Integration of multiple fingerprint
matching algorithms’, in Proceedings of the 5th Conference on Machine Learning and
Cybernetics, pp.3186–3189.
Zhang, J., Ou, Z. and Wei, H. (2006) ‘Fingerprint matching using phase-only correlation and
Fourier-Mellin transforms’, in Proceedings of the 6th International Conference on Intelligent
Systems Design and Applications, Vol. 2, No. 3, pp.379–383.
Zhu, Y–H., Jia, W. and Liu, L–F. (2009) ‘Palmprint recognition using band-limited phase-only
correlation and different representations’, in Proceedings of the 5th International Conference
on Intelligent Computing (ICIC 2009), Vol. 5754, pp.270–277.
ResearchGate has not been able to resolve any citations for this publication.
Full-text available
The bit-plane feature extraction approach has lately been introduced for face and palm-print recognition. This approach decomposes an 8-bit grey level image into eight groups of bit layers. The assumption of this approach is that the highest order of a bit-plane decomposition, which has the most significant bits of all pixels, contains the most biometric features. Nonetheless, most research has identified bit-plane images illustratively. Hence, in order to endorse the assumption, we performed an analysis on face and palm-print images to identify the bit-plane that contributes most significantly to the recognition performance. Analysis was done based on Principal Component Analysis (PCA). The first principal component was applied as it is defined for the largest possible variance of the data. Next, Euclidean distance was calculated for matching performance. It was observed that bit-plane 6 and 7 contributed significantly to recognition performance.
Full-text available
An 8-bit digital image consists of 256 levels of gray-value and 8 layers of multilevel information of bits known as bit-plane information. A novel method utilizing higher order bit-plane information that contains majority of visually significant data and dummy blank images as inputs to a multilayer feedforward Neural Network (NN) is proposed in this paper to perform face recognition. Experiments performed on the proposed face recognition model using two face databases, namely CMU AMP face expression database and Yale face database, show improvement in recognition rate compared to using only gray-level images as inputs to the NN.
Full-text available
SUMMARY This paper presents a technique for high-accuracy cor- respondence search between two images using Phase-Only Correlation (POC) and its performance evaluation in a D measurement application. The proposed technique employs (i) a coarse-to-fine strategy using image pyramids for correspondence search and (ii) a sub-pixel window align- ment technique for finding a pair of corresponding points with sub-pixel displacement accuracy. Experimental evaluation shows that the proposed method makes possible to estimate the displacement between correspond- ing points with approximately 0.05-pixel accuracy when using 11 × 11- pixel matching windows. This paper also describes an application of the proposed technique to passive D measurement system.
Full-text available
The Handbook of Information and Communication Security covers some of the latest advances in fundamentals, cryptography, intrusion detection, access control, networking (including extensive sections on optics and wireless systems), software, forensics, and legal issues. The editors intention, with respect to the presentation and sequencing of the chapters, was to create a reasonably natural flow between the various sub-topics. This handbook will be useful to researchers and graduate students in academia, as well as being an invaluable resource for university instructors who are searching for new material to cover in their security courses. In addition, the topics in this volume are highly relevant to the real world practice of information security, which should make this book a valuable resource for working IT professionals. This handbook will be a valuable resource for a diverse audience for many years to come.
A fingerprint matching algorithm compares two given fingerprints and returns either a degree of similarity (without loss of generality, a score between 0 and 1) or a binary decision (mated/non-mated). Only a few matching algorithms operate directly on grayscale fingerprint images; most of them require that an intermediate fingerprint representation be derived through a feature extraction stage (refer to Chapter 3). Without loss of generality, hereafter we denote the representation of the fingerprint acquired during enrollment as the template (T) and the representation of the fingerprint to be matched as the input (I). In case no feature extraction is performed, the fingerprint representation coincides with the grayscale fingerprint image itself; hence, throughout this chapter, we denote both raw fingerprint images and fingerprint feature vectors (e.g., minutiae) with T and I. The fingerprint feature extraction and matching algorithms are usually quite similar for both fingerprint verification and identification problems. This is because the fingerprint identification problem (i.e., searching for an input fingerprint in a database of N fingerprints) can be implemented as a sequential execution of N one-to-one comparisons (verifications) between pairs of fingerprints. The fingerprint classification and indexing techniques are usually exploited to speed up the search (refer to Chapter 5) in fingerprint identification problems.
A fingerprint is the reproduction of the exterior appearance of the fingertip epidermis. The most evident structural characteristic of a fingerprint is a pattern of interleaved ridges and valleys (Ashbaugh, 1999); in a fingerprint image, ridges (also called ridge lines) are dark whereas valleys are bright (see Figure 3.1). Ridges vary in width from 100 μm, for very thin ridges, to 300 μm for thick ridges. Generally, the period of a ridge/valley cycle is about 500 μm. Most injuries to a finger such as superficial burns, abrasions, or cuts do not affect the underlying ridge structure, and the original pattern is duplicated in any new skin that grows.
Conference Paper
Digital CT Scan images are exchanged among themedical practitioners for consultative purposes or discussing diagnostic and therapeutic measures. This exchange of medical images imposes two main constraints for medical information: (i) the Content Authentication and (ii) Copyright protection. To address these two issues, we have proposed a technique of watermarking for medical images which embeds two watermarks in the CT scan image. The binary pattern is embedded in region of interest (ROI) for content authentication and the composite watermark is embedded in the region of non interest (RONI) for copyright protection. Initially CT Scan image is divided in ROI and RONI by segmentation process. The segmented ROI contains some holes which are first filled. The image morphology is then used and closing operation is applied on the resultant image which eliminates small holes and fills gaps in the contour. Thus binary mask is obtained which is used to separate ROI from RONI. Before embedding the binary pattern in ROI for data authentication purpose, the Least Significant Bit (LSB) information of ROI is extracted first and is saved in separate store. Also, before embedding the composite watermark, pixels in RONI are scrambled for providing the further security. The payload in both ROI and RONI is embedded with simple LSB substitution technique. The extraction procedure is the reverse of embedding process. However few steps of extraction procedure are same as embedding procedure. We have used CT scan image database obtained from Radiology Department, Ackron University Ohio USA. The experimental results show that the proposed algorithm provides better security of medical information during transmission by addressing both issues of data authentication and copyright protection.
Conference Paper
This paper discusses about the effects of bit plane extraction in fingerprint recognition. An alternative approach to recognise a fingerprint from extracted bit plane is analysed in attempt to find the best bit plane used for recognition. An 8-bit greyscale fingerprint image is extracted into 8 different bit planes. Each bit plane of the images is then used as the input image for recognition. A fingerprint recognition algorithm using phase-only correlation (POC) is applied on the extracted bit planes. Based on the results of the analysis, the average recognition rate achieved in bit plane 7 is higher compared to the other bit planes. Three hundred samples of fingerprint images from FingerDOS are used for the experimental purposes.