Noname manuscript No.

(will be inserted by the editor)

Machine Assisted Authentication of Paper Currency: an
Experiment on Indian Banknotes
Ankush Roy Biswajit Halder Utpal Garain David S. Doermann
Received: date / Accepted: date
Abstract Automatic authentication of paper money
has been targeted. Indian bank notes are taken as ref-
erence to show how a system can be developed for dis-
criminating fake notes from genuine ones. Image pro-
cessing and pattern recognition techniques are used to
design the overall approach. The ability of the embed-
ded security aspects is thoroughly analysed for detect-
ing fake currencies. Real forensic samples are involved
in the experiment that shows a high precision machine
can be developed for authentication of paper money.
The system performance is reported in terms of both
accuracy and processing speed. Comparison with hu-
man subjects namely forensic experts and bank stas
clearly shows its applicability for mass checking of cur-
rency notes in the real world. The analysis of security
features to protect counterfeiting highlights some facts
that should be taken care of in future designing of cur-
rency notes.
A. Roy
Department of Computing Science
University of Alberta, Edmonton T6G 2E8, Canada
E-mail: [email protected]
B. Halder
Dept. of Computer Science
The University of Burdwan, Bardhaman, W.B., India
E-mail: [email protected]
U. Garain
CVPR Unit, Indian Statistical Unit
203 BT Road, Kolkata70018, India
E-mail: [email protected]
D.S. Doermann
Institute of Advanced Computer Studies
University of Maryland, College Park, USA
E-mail: [email protected]
1 Introduction
The problem of large scale counterfeiting paper cur-
rency poses a serious threat to our society as large
amount of fake notes causes economic instability. Coun-
terfeiting of currency notes aects the existence of the
monetary equilibrium as its value, velocity, output and
welfare may get aected. Most countries that use paper
currency for transactions are plagued by this problem.
Several media [1] [2] reports highlight the alarming rate
at which this counterfeits are increasing, the serious-
ness of the issue, as well as the continuous government
eorts to curb this problem [3]. Unfortunately counter-
feiters also adapt to the new security features that are
incorporated. Criminals continue to nd ways to repli-
cate the currency despite the new banknote security
features in place. There have been leaps and bounds in
the technical eld of counterfeit currencies, and this to-
gether with the recent advances in the digital scanning
and copying techniques has been an indomitable force.
This has moved the European Union legislators to draft
new guidelines so that computer and software manufac-
turers are to be forced to introduce new security mea-
sures to make it impossible for their products to be used
to copy banknotes [4]. However, it is almost impossible
to track and stop all of the counterfeiting eorts and
hence, we need to deploy better authentication systems
that carefully scrutinize notes before allowing them to
circulate.
The bank stas are specially trained to detect coun-
terfeit notes and so far they do it manually as the sort-
ing machines are not able to do this job. The technol-
ogy based on which sorting machines work is limited
to image based scanning, which is not foolproof when
it comes to detection of fake currencies. The technol-
ogy for sorting banknotes does not take into consid-
a
r
X
i
v
:
1
4
0
1
.
0
6
8
9
v
1


[
c
s
.
C
V
]


2

J
a
n

2
0
1
4
2 Ankush Roy et al.
eration the high intricacies involved into security fea-
tures, physical and chemical properties of papers, inks,
resins, chemical, etc. used to print the currency notes
and therefore, they are often unable to distinguish the
slight discrepancies between fake and genuine notes.
Sometimes the forensic experts (i.e. the questioned
document examiners) are involved to give their opin-
ions on suspected notes. Existing methods for detect-
ing and conrming the fake notes is too cumbersome
as this involves ling a case to the police, sending the
document for verication and then waiting for results
to come. On the other hand, the vast trac actually
calls for the introduction of machine authentication of
currency notes [5], [6]. Forensic agencies are always on
the lookout for systems that track these forgeries using
automated techniques and give a complete analysis of
the error report of the note in question. This analysis is
of particular interest to the community as they would
understand the robustness of the security features and
on which features to work on in future note design.
Further the speed introduced to the process due to au-
tomation will help in handling the huge volume of pa-
per currencies. Machine authentication technique could
also ensure that fraudulent samples are not passed via
ATMs and other cash delivery systems [7]. Even cur-
rency note counting or sorting machines may have such
equipment installed that looks to trap the counterfeit
currency note that might appear. This would trap the
fake notes that are circulated in the public directly by
the counterfeiters.
Our research is directed to this end. It attempts to
provide a complete automated approach for detection
of counterfeit currency notes. Also a thorough analy-
sis is provided that explores the performance of the
embedded security features and their sensitivity. This
analysis helps the regulatory bodies understand which
security feature are under what kind of threat of breach
and what modications could be done to improve the
design, making it less vulnerable to counterfeiting. De-
tailed experiments were done with real data to support
the claim. A comparative study is also reported, involv-
ing forensic document experts and bank sta to show
the applicability and robustness of the system at a grass
root level.
1.1 Previous Work
Research on manufacturing secure currency notes is in-
deed an old subject. Because of the commercial nature
of this research lots of research studies were patented
rather than published in scientic journals. U.S. Patent
issued in 1857 [8] may be the earliest attempt of an op-
tical method of manufacturing secure paper money. It
involved using paper tinted to absorb light, and print-
ing ink that also absorbed light rather than reecting
it so that clear photographic copies could not be made.
Many researchers assumed that the embedded water-
mark and security thread are hard to duplicate and
led patents on how to authenticate the watermark and
or security thread [9], [10] and [11]. The patent in [12]
described an approach that is based on the reective
property of the bank notes in question. A series of light
source is placed that provides wavelength of varying il-
lumination, which is used to measure the reective and
refractive response of the currency note images. This
data is then compared with the pre-calculated data on
original and fake notes to report the authenticity of the
notes. On the other hand, the patent in [13] proposed
a method to verify US currency notes by analysing dif-
ferent aspects of the ink used for printing.
Because of the commercial interest, the patents ex-
press little technical and experimental details and this
has restricted the research community to judge the per-
formance of the systems. On the contrary, the researchers
reporting details of their methods and experiments have
so far dealt with recognition of currency notes. They ad-
dress the problem of recognizing currency of dierent
countries (U.S. dollars, Euro notes, etc.), and dierent
denominations for a given currency. For this purpose,
Neural Networks [4], [8], [9], [14], [15], and [16]; Genetic
Algorithm [17] and Hidden Markov Model [18] have so
far been used. These studies have successfully addressed
the currency or denomination recognition problem but
did not consider whether the input bank note is gen-
uine or fake. There are a few studies which address this
problem and give details about their methods and ex-
periments. The paper in [19] is an example. In this ar-
ticle, the authors proposed a semi-automatic approach
for characterizing and distinguishing original and fake
Euro notes. Their method is based on the analysis of
several areas of the banknotes using a Fourier trans-
formed infra-red spectrometer with a microscope with
an attenuated total reectance (ATR) objective. They
considered four dierent regions of a note and observed
that fake notes are easily identiable from the anal-
ysis of the spectra corresponding to the four regions.
However, the authors did not propose any automated
scheme for authentication.
Later on, the authors in [20] describe another system
for authenticating Bangladeshi Bank Notes. They as-
sume that original currencies under test have the bank
name printed in micro letter print. They scan this part
(the region where the bank name should be) using a
grid scanner and the textual images are fed in an op-
tical character recognition engine that matches charac-
ters with prototypes. Since the fake currencies are as-
Machine Assisted Authentication of Paper Currency: an Experiment on Indian Banknotes 3
sumed not to have the text they show very low matching
percentage. The algorithm is heavily dependent on one
feature which makes the system very sensitive. The sys-
tem would fail miserable if the counterfeiters happen to
develop means of duplicating the feature in question.
Very recently, a commercial system [21] is available
in the market for detecting fake notes. In order to au-
thenticate a paper currency, the system makes use of
several hardware components like light intensity com-
pensation circuits, magnetic level circuits, visual light
sensors, Infra-Red ray sensors, Ultraviolet Ray Sensors
and many other technical light sensors. These hardware
components function simultaneously and compare cru-
cial security features like ones from the based materials
like paper, ink, security thread, uorescent resins, mi-
cro prints, secret watermarks, intaglio patterns, etc.,
against the data stored in standard memory bank. The
memory bank keeps information about the characteris-
tics of genuine and counterfeit notes. For a given note, if
the features match to those of the real ones, the system
passes the note as genuine. If any dierence is spotted,
the machine saves the features match those of counter-
feits and these notes are rejected. In case of no match
either with genuine note or with counterfeit notes, the
notes should be identied as genuine but worn out and
hence machine passes it as genuine in case of normally
worn out notes.
1.2 Contribution of our Work
After reviewing the literature we nd that there have
been little studies reporting technical and experimental
details on how to automatically authenticate currency
notes. Many studies rely on one or two features that also
can be duplicated using todays high end technology.
As there are so many security features in the currency
notes, analysing only a certain aspect of the note secu-
rity may not be a good choice. A complete integrated
framework has been missing that looks into many as-
pects like security features in printing, ink, background
artwork, watermark, security thread, etc. Another gen-
eral shortcoming in the existing studies is the use of syn-
thetic data. Many authors generate samples at lab to
test their algorithms. Therefore, performances of these
algorithms on real forensic samples are yet to be ex-
plored.
Earlier, we provided conclusive evidence of decision
making using printing techniques [22], but the approach
looked only on the printing process and dealt with a
small database to test the initial system performance.
In this paper, we have incorporated several other se-
curity aspects based on ink, security thread and an
art work based feature set that supplements the print-
ing style based approach. A large dataset comprising
of 1000 notes (500 for each genuine (G) and fake (F)
class) has been used to test the system performance and
a thorough analysis is provided by comparison with real
forensic document examiners to show the applicability
of our research in the real world scenario. Analysis is
also done to report the speed of the system. Popular
pattern recognition tools like the k-means, neural net-
work and support vector machines are used to demon-
strate the system performance. We have also shown the
robustness of each feature in tackling the problem and
thus pointing out the sensitivity of the features. The
feature level analysis brings out important views that
the note designing agencies could keep in mind regard-
ing the susceptibility of the features to be duplicated. A
sequential ordering of security features is also suggested
that should be maintained while testing to maximize
the performance accuracy. Involvement of real foren-
sic samples is another signicant aspect of this study.
Indian bank notes are taken as a reference. In fact, be-
cause of highly heterogeneous concepts involved with
Indian currencies, authentication of these notes poses a
big technical challenge.
2 Security features of Indian Bank Notes
The impact of the document and its mobility as a trans-
action medium determines the amount of security in
place for its protection. There have been a lot of re-
search studies for providing security in paper docu-
ments [23]. Bank notes being the principal monetary
exchange medium do have a lot of security features to
prevent counterfeit [24]. The Apex bank of every coun-
try is responsible for placing these security features to
prevent forgery. In India it is the responsibility of the
Reserve Bank of India (RBI) [25]. The security features
in a currency note are mainly on its paper, design and
printing. Authentication of currency notes is thus re-
fer to authentication of the following areas: i) Currency
notepaper, ii) Currency note printing technique, iii) Ink
used for currency note printing, iv) Currency note de-
sign, v) Other security features (e.g. the thread, the reg-
istration mark, and many others) that are intentionally
incorporated to check the authenticity. These security
features provide a tough challenge to the counterfeit-
ers who attempt to replicate them. Important security
features are associated with the currency notepaper it-
self. Physical features of currency note are based on
its cut size of length, width, grammage and thickness
of paper. The paper has a unique feel, crackling sound
and it is constituted of high quality with 100% cot-
ton or wood pulp lending a particular color, a unique
4 Ankush Roy et al.
ber length, surface nish, a typical opacity, and its
capacity of extra strength of folding. Watermarks and
security thread are another important parts of security
aspect of paper money. Important property of water-
mark is that it cannot be replicated on scanning or
by photocopy equipment. Examination of watermarks
checks its design, size and thickness using transparent
light. The security thread appears to the left of the Ma-
hatma Gandhi portrait is partially embedded and par-
tially visible. On seeing this thread with an ultraviolet
light exposure the thread appears in a single line. This
is also a proof of the registration of the note that both
the sides of the note are properly aligned. This thread
has the writings of RBI and Bharata (India) in De-
vanagari) alternatively written on it. Denomination of a
note (e.g. 500, or 1000) is also embedded in the thread.
Fig. 1 shows the signicant security features embedded
in both sides of a 500 Rupee note (source: Reserve bank
of India).
(a)
(b)
Fig. 1 Signicant security features in Indian banknote: (a)
Front side features: (i) Multidirectional artisan lines, (ii) In-
taglio printing, (iii) Omron, (iv) Micro text, (v) Latent im-
age, (vi) Blind mark, (vii) See through, (viii) Fluorescent ink
numbers, (ix) Optically variable ink, (x) Security thread with
clear text, and (xi) Hand graved portrait; (b) Back side fea-
tures: (i) Multi-directional artisan lines, (ii) Intaglio printing,
(iii) Gandhi water mark, (iv) See through, and (v) Omron
features.
A lot of important security features are involved in
the design of currency note. The Guilloche design, por-
trait design, micro lettering, type face, font size, color,
see through register, anti-scan lines, Braille mark, rain-
bow eect, layers of CYMK, bleeding eects, latent im-
age eects, etc. all are involved as part of security de-
sign. Another level of distinction of currency notes is
the printing process. Among the numerous printing pro-
cesses that can be used for printing currency note only
a few processes are used for printing of Indian currency
notes. Among them intaglio printing is of prime impor-
tance because of the typical signature it leave and it be-
ing hard to duplicate. The basic components of printing
ink are pigment, solvent, and drier. Where ink pigment
is responsible for color eect on substrate, drier is re-
sponsible to bind ink pigment to substrate and solvent
is responsible for solubility of ink pigment and drier. Ex-
amination of nal printing eect is an important aspect
for verication of security document authenticity. Final
eects on currency note are dened by unique color, im-
pression, water resistance, line work (width, thickness,
sharpness, etc.), halftone eect, digitized patterns and
also reectivity and feel/tactility.
3 Overview of the proposed method
The proposed method is based on image processing [26]
and pattern recognition principles [27]. The feature ex-
traction in this experiment is largely dominated by the
input from the forensic experts making sure that every
aspect of the security features is considered when choos-
ing features. As all the features used by the experts [28]
cannot be captured computationally, a subset of the fea-
tures is used. Some new features which are eective for
detecting fake notes, but are dicult to check manu-
ally, are also added. The feature space is analysed and
visualized by using clustering technique. The decision
making process is built using two dierent classiers: i)
Articial Neural Networks (ANNs) ii) Support Vector
Machines. Furthermore, a Linear Discriminate Analysis
is used to measure the performance of each feature. Our
feature extraction process considers four dierent secu-
rity aspects of the banknote: i) printing technique, ii)
Ink property, iii) Thread and (iv) the Art work used in
designing the note. The features and rationales behind
choosing them are explained below.
3.1 Preprocessing
Dierent security features are available at dierent parts
of the banknote image. So the initial scanned image
needs to be divided into distinct ROIs. The image of
the currency note is registered using Hough Transform
on a Canny edge detected image (Fig. 2(a)). Template
matching of the denomination (500/1000) and the
Machine Assisted Authentication of Paper Currency: an Experiment on Indian Banknotes 5
Mahatma Gandhi portrait generates two xed positions
as reference around which the rest of the ROIs are ex-
tracted (Fig. 2(b)).
(a)
(b)
Fig. 2 Preprocessing steps: (a) Image registered using
Hough Transform (b) Extracted ROI (i) matched por-
trait(red)(ii)matched denomination(red) (iii) ROI from la-
tent image(blue) (iv) microprint lines(blue) (v) Intaglio
print(blue) (vi) Central pattern(blue) and (vii) Security
thread(blue).
The horizontal strip just above the registered por-
trait of Gandhi is used to segment the intaglio fonts.
Text extraction from this part is done using the verti-
cal and horizontal pixel projection techniques (Fig. 3).
The biggest black pixel blob in the image is the area to
be focussed for the extracting features using latent scan.
A vertical strip just beside the denomination is used for
the security thread based measures and nally the re-
gion between the registered Mahatma Gandhi portrait
and the largest black blob is used for the micro-print
line features.
Fig. 3 Text extraction using pixel projection: (top) Black
pixel projection on x-axis, (bottom) the horizontal strip used
to separate Intaglio fonts in English.
3.2 Features
3.2.1 Printing Technique
Intaglio printing is used for printing currency notes in
India. The denomination of the note and Reserve Bank
of India are printed on the face of the notes and are al-
ways printed using the Intaglio method. This method of
printing leaves several signatures that are hard to repli-
cate [28]. We have analysed some of the features and
tried to dierentiate the fake notes from the genuine
ones based on printing technique detection. Following
are the features used in detecting printing technique.
Dominant intensity (f
1
p
) is used to capture slight dif-
ference in brightness (or glossiness) of banknotes. Math-
ematically this is represented as follows,
f
1
p
= x : f(x) = max(intensity histogram). (1)
The feature, Hole count (f
2
p
) checks the textural
similarity of the printed character strokes in a note
by counting the number of eight connected white pixel
cluster (dened as hole) divided by the area of the char-
acter stroke as shown in Eq.(2),
f
2
p
=
#8 connected white patch in char strokes
Area of character stroke
. (2)
Average hue (f
3
p
) gives an assessment of the quality
of color. This feature is computed in HSV space on the
Hue (H) stream as follows,
f
3
p
= Average(H). (3)
R.M.S. contrast (f
4
p
) in Eq.(4) measures the vary-
ing dierence in brightness of the two classes (Gen-
uine/Fake) of notes. Mathematically it is expressed as
follows (where I
i
and

I denote the intensity of the i-th
pixel and the mean intensity, respectively),
f
4
p
=

1
N 1
N

i=1
(I
i


I)
2
. (4)
The key tone (f
5
p
) value gives us the information
about the intensity zone where most of the information
is stored by calculating the mean of the intensity prole
of the character stroke after masking as given below,
Image Masked character f
5
p
= Mean(Intensity).(5)
Average color (f
6
p
) assesses a reconstituted color ma-
trix based on a scalar parameter (p). This checks the
color composition of the print according to the note
issuing authority of India. For instance, the principal
streams in Intaglio character stroke are blue and black
in 500 denominations. The average color is computed
as below,
f
6
p
=

s(i)
N
(6)
6 Ankush Roy et al.
where s(i) is dened as
s(i) = pB
blue
(i) + (1 p)B
black
(i), 0 < p < 0.5
and B
blue
and B
black
correspond to blue and black
strokes, respectively.
Along with these six features, three other features
are extracted: edge roughness E
PBER
, (f
7
p
) (Eq. 7),
area dierence (f
8
p
) (Eq. 8) and correlation coecient
(f
9
p
). These features are computed based on the work of
Breuel et al. [29], [30]. The edge roughness is computed
as
f
7
p
= E
PBER
= (p
a
p
b
)/p
b
(7)
where p
a
is the perimeter of the original image, p
b
is perimeter of the ltered (median lter) binary image
and E
PBER
is the perimeter based edge roughness. In
calculating this feature, the character image is rst bi-
narized using Otsu threshold value (say, T) (A
otsu
) and
then the same image is again binarized using a dierent
threshold value that is calculated by adding a normal-
ized parameter sc to T (A
otsu+sc
). The area dierence
is computed as
f
8
p
= Area Dierence =
|A
otsu+sc
A
otsu
|
A
otsu
. (8)
The correlation coecient is computed as
f
9
p
=

(i,j)ROI
(A(i, j)

A)(B(i, j)

B)

(i,j)
(A(i, j)

A)
2

(i,j)
(B(i, j)

B)
2
(9)
where (i, j) ROI, A is the original gray value
image, B is the corresponding binary image,

A and

B
are the mean of A and B, respectively.
Based on the above nine features, classiers are trained
to identify fake notes based on printing technique.
3.2.2 Ink Properties
The reaction of the ink on a particular substrate is dif-
ferent for dierent inks. This dierence actually lends a
typical signature indicating the authenticity of a note.
CCRatio (f
1
i
) Colour composition of the central zone
(Fig. 4) of a note is analysed by doing an independent
component analysis. This was followed by a ltration
method that keeps those pixels ON where the green
component index in the RGB color space is higher than
the blue component index which is in turn higher than
the red component index to generate a mask. Compu-
tationally it is represented by color composition ratio
(CCRatio) feature which is dened as
f
1
i
=
#ON Pixel in mask
#Pixels in mask
. (10)
The number of pixels are xed as the images are regis-
tered using a 4-point registration prior to processing.
(a) (b)
(c) (d)
Fig. 4 Analysis of colour composition: (a) and (c) UV scans
of fake and original image; (b) and (d) are resultant images
from the left hand side counterparts after ltration to check
colour composition.
Micro letter (f
2
i
) This feature appears between the ver-
tical band and Mahatma Gandhi portrait in the notes.
In notes of denominations 20 and above, the denomina-
tional value and RBI constitute the micro letters. In
our study, we have looked into the colour of these micro
letters. The RGB values are rst transformed to a spe-
cic absolute colour space. This adjustment makes the
resulting data device independent. The masked image
was changed from RGB to L

colour space using

the CEILAB Illuminant D65 as a reference [31]. The re-
sultant distribution of b

index is plotted in Fig. 5. The

dierence (visually) between genuine and fake notes is
seen. Computationally, spread of the index distribution
is captured as the feature f
2
i
by calculating the standard
deviation (spread) values of the b

index distribution as
formulated below,
f
2
i
= Spread =

N
i=1
b

i


b

N 1
. (11)
Ink Fluidity (f
3
i
) It is observed that the ink used to
print genuine currency notes blot considerably greater
than the counterfeit ink. The study of uidity of ink as
a vision based feature was done by K. Franke et al. [32].
Following this study we developed a feature that would
computationally help in decision making about the ink
Machine Assisted Authentication of Paper Currency: an Experiment on Indian Banknotes 7
(a) (b)
(c)
Fig. 5 Analysis of micro lettering: (a) genuine banknote, (b)
fake banknote; (c) b

stream index plot of genuine (blue line)

and fake (red line).
authenticity. Edges of the print were taken and the in-
tensity prole was plotted (Fig. 6). We then normalise
the curve using an averaging kernel. A steady value is
computed in Eq. 12 as follows. Let f(x) be the number
of pixels having intensity x, x X = {x
1
, x
2
...x
n
},
where x
1
refers to the intensity value corresponding to
maximum pixel count and x
k
= x
k1
+ 1. We dene
f(x) as f(x) = f(x
k+1
) f(x
k
), 0 < k < n 1
where x
n
> x
n1
> x
n2
> ... > x
2
> x
1
.
The f(x) vector would always start with a nega-
tive quantity as the very rst value in this vector is the
dierence from the highest pixel count. The rst posi-
tive entry is found (x
p
) and the steady value is gener-
ated as follows,
steady =

n
p
f(x)
n + 1 p
(12)
where p is the position of the 1st positive entry.
The percentage overshoot is then recorded as a fea-
ture using the steady value in Eq.(13) and computed as
follows,
f
3
i
=
overshoot
steady
=
max
steady
1. (13)
3.2.3 Thread
Two security thread related features are considered: the
registration of the notes and the text in the security
strip.
Registration (f
1
t
) The thread should always appear as
a single line. This is a way to test of the registration
of the notes. We check this using a binary feature, f
1
t
,
(a) (b)
(c) (d)
(e) (f)
Fig. 6 Ink analysis: (a) Genuine ink spread, (b) Fake ink
spread, (c) Histogram of genuine ink, (d), Histogram of fake
ink, (e) Normalized graph with steady value (genuine ink),
and (f ) Normalized graph with steady value (fake ink).
which decides whether a note is genuine (G = 1) or
fake (D = 0). Two sets of thick blobs are found (refer
Fig. 7(i)b), one represents the thread parts seen from
the front while the other represents the thread parts on
the back of the note. Two lines, one for the front and the
other for the back, are t through the centroid (c
x
,c
y
)
points of the corresponding blobs. The centroid of a
blob is calculated using the following formulae where A
be the area of the blob.
c
x
=

n1
i=0
(x
i
+x
i+1
)(x
i
y
i+1
y
i
x
i+1
)
6A
c
y
=

n1
i=0
(y
i
+y
i+1
)(x
i
y
i+1
y
i
x
i+1
)
6A
This is followed by a distance check of the lines using
a threshold distance t, empirically computed from 100
note samples.
d(t
f
, t
b
) < t, Genuine,
>= t, Fake.
(14)
where t
f
= foreground line points, t
b
= background line
points.
The feature, f
1
t
is a binary feature which generates
a decision based on Eq. 14. Fig. 7(i) shows the registra-
tion problem of thread in a fake note.
8 Ankush Roy et al.
Text in Thread (f
2
t
) This is another binary feature that
checks whether thread texts exist. The texts RBI and
Bharat (Hindi for India) in Devanagari script are
written on original notes where these two words appear
alternatively. We extracted the text portion from the
threads and then used conventional pattern matching
tools to compare. There were only 4 such texts patterns
as shown in Fig. 7(ii)(a-d) to compare so the templates
were extracted from the original image and then used
to as ground truth data for pattern matching. Majority
of fake notes showed negligible matching because they
do not have any text as shown in Fig. 7(ii)(f).
(i)
(ii)
Fig. 7 Analysis of security thread: (i) Security Thread: (a)
fake note image (b) thick blobs representing the thread on
the front (c) line in front (d) two lines which do not overlap;
(ii) Text in Security Thread: (a)-(d) four occurring patterns
(e) original note (f) fake note.
3.2.4 Art Work
This section deals with printing patterns that are intri-
cately introduced in note design to prevent the coun-
terfeiters from replicating them. Initially the image is
passed through a median (3 3 sub window) lter to
remove impulsive noise. Next, the centroid of each dot
is mapped as shown in Fig. 8. The three features de-
scribed below are extracted and analysed.
Fig. 8 Analysis of dot distribution: (a) Genuine note (b) dis-
tribution of dots centroids for the genuine note (c) Fake note
(d) distribution of dot centroids for the fake note.
Dot distribution (f
1
a
) The distribution of the dot cen-
troids gives us the impression that in the fake note the
distribution of the dots are far less uniform when com-
pared to the genuine notes. Entropy count provides a
measure of this randomness. The entropy (H) is calcu-
lated as a feature, f
1
a
, and the following equation mea-
sures it as
f
1
a
= H =
n

i=1
p(x
i
) log p(x
i
) (15)
where n is the number of dots.
Cluster distribution and dot density features (f
2
a
and
f
3
a
) We also compute the number of clusters occur-
ring at the character strokes of the letters. An unsu-
pervised agglomerative hierarchical clustering scheme
is used with a Euclidean distance check to calculate the
number of clusters. Let C
1
and C
2
be the two clus-
ters and d is the distance matrix, then max{d(x, y) :
x C
1
, y C
2
}. An iterative process continues until
the separation of the clusters exceeds a threshold (in-
dicated by a Euclidean distance). Cluster density f
2
a
is
dened as
f
2
a
=
#Clusters
Area of Character Stroke
. (16)
From the Fig. 8 it is evident that the original notes have
a more even distribution of dots and far more dots in
the character stroke. This dierence in the dot count
is used as another simple feature which gives the dot
density, f
3
a
as,
f
3
a
=
#dots
Area of Character Stroke
. (17)
Machine Assisted Authentication of Paper Currency: an Experiment on Indian Banknotes 9
Adaptive Otsu threshold removes most of the back-
ground leaving only the character stroke whose area
is calculated.
Reading Latent Denomination (f
4
a
) When viewing the
strip left to the Mahatma Gandhis picture, the denom-
ination of the note is seen engraved quite distinctly.
The machine readability of these denomination digits
is measured in Eq. 18. This zone comprises of two sets
of lines- horizontal and vertical. The vertical lines repre-
sent the text part. The sharpness of the lines is dierent
in the two categories of notes (genuine vs. fake) lending
them dierent readability index f
4
a
which is dened as
f
4
a
= Score of NN classier. (18)
An MLP-NN classier was trained with digit sam-
ples taken from the genuine notes. The vertical lines
need to be extracted for the text to be read properly so
we use a 3 3 convolution matrix (CM)
CM =

2 4 2
2 4 2
2 4 2

that reduces the number of horizontal lines in the image

and strengthens the vertical lines.
The image that results from this operation has a
lot of clutter and noise (Fig. 10(b)). A particular com-
ponent tag is assigned to each 8-connected cluster and
the number of pixels in each cluster is recorded. Fig. 9
shows the histogram of component size vs. component
tag. A threshold is selected from this histogram to elim-
inate most of the clutter. This method brings out the
latent information from the image. Fig. 10(c) shows the
resultant image after thresholding.
Fig. 9 Component size (along the y-axis) vs. Component tag
(along the x-axis).
4 Experiment
Experiment is conducted to check the validity of the
proposed approach. Real life forensic samples are con-
(a)
(b)
(c)
Fig. 10 Analysis of latent image: (a) Image of genuine note,
(b) after convolution, (c) after ltration.
sidered in this experiment. This section explains the
data set, the capturing conditions, our experimental
strategies, and nally, the results.
4.1 Dataset
This work considers the real data that was collected
from the forensic department of India. The dataset has
a total of 500 original and 500 fake note images scanned
into jpeg format. The scanning involves dierent medium
as explained in the next section. Forensic experts re-
ceived the fake notes at dierent times and from dif-
ferent groups. From their experience they note that a
particular group shows a particular kind of prociency
making fake notes. Therefore, only few fake note sam-
ples were considered for a particular group. Fake note
samples from any such groups are considered so that
samples show a wide variety and diversity in counter-
feiting features. Two denominations namely 500 and
1000 (1000 being the highest denomination for Indian
bank notes) are considered as fake note makers mostly
target these two denominations. The experts encounter
the highest number of fake notes for denomination of
10 Ankush Roy et al.
Table 1 Capturing Conditions
Condition UV Scan Latent Flood
I II
Light Source 365nm UV Co-axial 100% 100%
Long pass VIS VIS VIS VIS
Band pass OFF OFF OFF OFF
Magnitude 2.069 9.64 29.97 40.08
Gain 12dB 1 dB 0 dB 2 dB
Iris 73% 50% 64% 64%
Integration
1
4
1
8
1
8
1
8
500. The dataset also considers 500 and 1000 denomi-
nations in a ratio of 7:3.
The entire dataset is divided into two sets. The Set
1 consists of 500 fake notes and 500 original notes. This
was used to train, test and validate the system perfor-
mance and analyse the accuracy of the features. Set 2
consisting of 200 fake notes and 200 original notes ran-
domly selected from the Set 1 and were used to test
the system in comparison to forensic experts and bank
sta. Since forensic experts were involved the set was
deliberately kept small as it took quite some time for
them to deliver results by following the existing manual
process.
4.2 Capturing conditions
In our experiment, four dierent scanning techniques:
(i) UV Scan, (ii) Latent, (iii) Flood I and (iv) Flood II,
are used to get image of a currency note. Dierent re-
gions of a note are scanned under dierent techniques.
For example, the central design of the note appears
prominent under ultraviolet (UV) ray, hence, this re-
gion is scanned with UV ray. There are latent marks on
the side of a note. The parameters of the cameras used
for imaging are given in Table 1. As explained in Sec-
tion 3, the proposed method captures several features
in order to authenticate a note. All these features are
not extracted from the same type of scanning medium.
Four dierent scans give dierent features as listed in
Table 2. It is noted that while scanning dierent re-
gions of a note under dierent capturing conditions, we
have used only one wave length of the light source. This
would lead to a relatively low cost adaptation of our
proposed method as most of the commercial systems
and patents (as discussed in Sec. 1) have extensively
used high cost multiple light sources of dierent wave-
lengths.
4.3 Experimental Strategies
The authentication of notes is modeled as a 2-class clas-
sication problem: discrimination of genuine (denoted
Table 2 Source of features
Features Captured from
I
N
K
f
e
a
t
u
r
e
s
CC
ratio
, f
1
i
UV Scan
Micro letter, f
2
i
Flood I
Ink uidity, f
3
i
Flood II
T
h
r
e
a
d
f
e
a
t
u
r
e
sRegistration, f
1
t
UV Scan
Thread text, f
2
t
UV Scan
A
r
t
w
o
r
k
f
e
a
t
u
r
e
sDot distribution, f
1
a
Flood I
Cluster distribution, f
2
a
Flood I
Dot density, f
3
a
Flood I
Latent denomination, f
4
a
Latent
P
r
i
n
t
i
n
g
f
e
a
t
u
r
e
s
All the nine features related Flood II
to printing techniques
by G) vs. fake (denoted by F). Dierent groups of fea-
tures are taken into account to accomplish this classi-
cation task. The feature groups, as discussed in the
previous section, are based on Ink, Thread, Artwork
and Printing technique. Each feature group is used in
a bi-clustering of the feature vectors to visualize their
distribution in the feature space and then again for clas-
sifying the samples by Support Vector Machines (SVM)
and Articial Neural Network (ANN) scheme. Four-fold
cross validation method is followed using both the SVM
and ANN based classication. The data set is divided
into 2:1:1 ratio to generate training, validation and test
sets. The relative robustness of the individual features
is measured and Linear Discriminant Analysis (LDA)
is used to provide a comparative performance analysis
of the feature groups.
Classication with SVM makes use two kernels (poly-
nomial kernel and RBF kernel). Mean squared error
(MSE) is computed as follows: MSE =

T
j=1

V
i=1
(d
ij
y
ij
)
2
V T
,
where V and T are the number of support vectors and
test samples, respectively, d
ij
and y
ij
are the desired
output and the SVM output, respectively for the i-
th support vector and the j-th test sample. The neu-
ral network used for classication consists of three in-
put nodes and 2 output or decision nodes with 1 hid-
den layer. Normal back propagation algorithm has been
used for training with an activation function given by
f(x) =
e
x
1+e
x
. A gradient descent method is used to
nd the optimized set of connection weights that are
updated as per the following equation,
W
t+1
= W
t
+ (
E
W
)|
W(t)
+ (W
t
W
t1
). (19)
Next, we use Fisher linear discrimination analysis
(LDA) [33] to study the performance of individual fea-
tures. Suppose is the normal to the discriminating
hyper plane,
y=0
,
y=1
, are the sample means,

y=0
and

y=1
are the covariance matrices of the two classes
(genuine and fake). Now we can consider that for one
Machine Assisted Authentication of Paper Currency: an Experiment on Indian Banknotes 11
class, y = 0 for
T
.X + b 0 and for the other class,
y = 1 for
T
.X +b < 0. From these equations, param-
eter vector is computed to maximize class separabil-
ity criterion and b is the bias, which lies in between the
means of the training samples projected onto this direc-
tion. The separation between these two distributions is
to be the ratio of the variance between the two classes
and is given by
S =

2
between

2
within
=
(.
y=1
.
y=0
)
2
(
T

y=1
+
T

y=0
)
. (20)
For Fisher LDA, this separation achieves maximum
when
w = (

y=1
+

y=0
)
1
(
y=1

y=0
). (21)
Finally, the cumulative accuracy taking all the features
into consideration is computed and a study is conducted
to nd out a proper sequence in which the features are
to be tested. This is important as this reduces the load
of the machine if at various stages the number of notes
to be checked could be reduced without sacricing the
accuracy.
4.4 Experimental Results
At rst, results are reported based on individual feature
groups. The capacity of these features are then analysed
in the context of detecting fake notes. Finally, results
are reported considering all the features and a sequence
by which features are to be checked is highlighted in
order to authenticate a bank note in question.
Accuracy of Ink Based Features. There are three ink
based features: f
1
i
,f
2
i
and f
3
i
. The bi-clustering results
on Set 1 using these three features are shown in Ta-
ble 3. Table 4 reports the classication results when
SVM and ANN, respectively are used as classiers.
Table 3 Bi-Clustering using Ink Based Features
#Samples of #Sample of Clustering
Genuine (G) Fake (F) Accuracy
G D G D
Iteration1 478 11 22 489 967 (96.7%)
Iteration2 471 18 29 482 953 (95.3%)
Iteration3 464 22 36 478 942 (94.2%)
Avg. 95.4%
Accuracy of Art Work Based Features. There are four
features which are captured from the note artwork: f
1
a
,f
2
a
,
Table 4 Ink Features based classication of Currency Notes
using SVM and ANN on Test set
SVM ANN
%Correct MSE
Poly RBF Poly RBF
Fold1 100.0 94.0 0.491 0.530 100.00%
Fold2 98.0 94.0 0.217 0.219 99.30%
Fold3 98.0 92.0 0.103 0.199 98.50%
Fold4 98.0 98.0 0.064 0.256 96.00%
Avg. 98.5 94.5 0.219 0.301 98.45%
f
3
a
and f
4
a
as discussed in Section-3. Bi-clustering re-
sults using these features are shown in Table 5. Clas-
sication accuracies by SVM and ANN are reported in
Table 6.
Table 5 Bi-Clustering using Artwork Based Features
#Samples of #Sample of Clustering
Genuine (G) Fake (F) Accuracy
G D G D
Iteration1 476 15 24 485 961 (96.10%)
Iteration2 474 14 26 486 960 (96.00%)
Iteration3 478 18 22 482 960 (96.00%)
Avg. 96.0%
Table 6 Art work based classication of Currency Notes us-
ing SVM and ANN on Test set
SVM ANN
%Correct MSE
Poly RBF Poly RBF
Fold1 99.00 98.00 0.011721 0.00527 100.0%
Fold2 100.00 100.00 0.018473 0.008652 98.0%
Fold3 100.00 99.00 0.012698 0.006285 99.0%
Fold4 98.00 99.00 0.010526 0.008807 99.0%
Avg. 99.25 99.00 0.0133545 0.007253 99.0%
Accuracy of printing based features. A comprehensive
study of the printing based features (f
1
p
-f
9
p
) is presented
in Table 7.
Table 7 Printing Technique based classication of Currency
Notes using SVM and ANN on test set
SVM ANN
%Correct MSE
Poly RBF Poly RBF
Fold1 100 100.00 0.000121 0.00527 100.00%
Fold2 100 100.00 0.000873 0.008652 99.30%
Fold3 100 100.00 0.01028 0.006285 98.50%
Fold4 100 99.00 0.000205 0.008807 98.00%
Avg. 100 99.75 0.002869 0.007253 98.95%
12 Ankush Roy et al.
In spite of the high accuracy of the individual fea-
ture group more than one feature group makes the sys-
tem more robust. This makes it more immune to the
counterfeiters eort, even if the fake note generators
happen to develop means to surpass the security mea-
sures of any particular feature set. Use of other features
would help in maintaining the high performance of the
system.
With currency notes we are extra careful in mak-
ing the system conservative so that no fake currency
passes through the system. The two thread based fea-
tures (f
1
t
,f
2
t
) are therefore used upfront that are bi-
nary in nature and discriminate notes based on either
the presence or absence of two features (namely regis-
tration (f
1
t
) and text(f
2
2
) as dened in Sec.3.2.3) into
Genuine (G) and Fake (F). This reduces the processing
time too which we have analysed next.
Recognition accuracy of the complete system. The per-
formance of individual feature group is quite good but
not perfect. Now all the features are taken together to
test the system. Individual feature groups are selected
in dierent sequence in order to optimize the perfor-
mance of the system in terms of accuracy, speed and
computational overhead. The Table 8 shows that the
system operates fastest if the thread based features is
placed in the beginning. A total of 1000 notes (500:G,
500:F ) were used to compute the speed of the system.
The time reported here comprises the time required to
scan the notes using the four scanning methods men-
tioned in Table 1, for registration to gure out which
part of the image should be fed to which module and
nally to execute the authentication framework. With
an improved scanner this time would certainly go down.
Table 8 Processing time under dierent ordering of security
features
Ordering of Test Set Accuracy Time
features G:Genuine (Mins)
F:Fake
1.Thread (F:500 100% 160
2.Ink G:500)
3.Printing Technique
4.Art Work
1.Ink (F:500 100% 222
2.Thread G:500)
3.Printing Technique
4.Art Work
1.Ink (F:500 100% 271
2.Printing Technique G:500)
3.Thread
4.Art Work
1.Ink (F:500 100% 324
2.Printing Technique G:500)
3.Art Work
4.Thread
It is also noted that all the features are needed to
make the system as accurate as possible. The nal sys-
tem integrates the two classiers (SVM and ANN) that
are used for discrimination between genuine and fake
currencies. A majority vote approach is followed in in-
tegrating results from these two classiers. A note is
rejected if a tie occurs during majority voting. This
makes the system conservative but reduces the risk of
accepting a fake sample as genuine.
Comparison with Respect to Human Experts. We con-
ducted a comparative study to check the systems per-
formance with respect to human experts. Two cate-
gories of people were involved in this experiment. The
rst group consisted of trained forensic experts and
the other group referred to bank employee who were
the rst in line to receive the bank notes from circula-
tion. The result is tabulated in Table 9. At every run
the notes were selected at random from Set 2 of the
database as mentioned in Sec. 4.1.
Forensic experts analyse each note individually us-
ing various optoelectronic devices which takes quite some
time. So, the dataset was kept deliberately small to ac-
count for this long process. The bank sta has a rea-
sonable speed of producing results but there accuracy
suers mainly because of the qualitative aspect involved
in the authentication process. Bank stas generally au-
thenticate notes by feeling the currency paper and quick
checking the note under light. They go for further ver-
ication only if any suspicious note is encounter but
then again the lack of sophisticated machines hampers
the overall accuracy.
Table 9 Performance with respect to human subjects (accu-
racies are reported for detection of fake notes).
Testing Samples Test Accuracy Time
Method Result (%) (min)
Forensic F:50, G:25 F:50, G:25 100 124
Experts F:50, G:50 F:50, G:50 100 166
F:25, G:100 F:25, G:100 100 204
Bank F:50, G:25 F:46, G:25 92 28
Employees F:50, G:50 F:47, G:50 94 36
F:25, G:100 F:23, G:100 92 41
Our F:50, G:25 F:50, G:25 100 12
System F:50, G:50 F:50, G:49 100 16
F:25, G:100 F:25, G:97 100 20
Relative performance of the feature group The ability
of the three feature groups, namely Ink, Artwork and
Printing technique based features, in detecting fake cur-
rency notes is analysed using Fisher linear discrimina-
tion analysis (LDA) [33]. The previous sub-section (4.3)
Machine Assisted Authentication of Paper Currency: an Experiment on Indian Banknotes 13
highlights how we use LDA for this purpose. The pro-
jection of the individual feature groups are taken on the
best discriminant plane and further mapped to show the
separability of the feature groups in a 2-D plot. Fig. 11
shows the results. Finally, performance of the individ-
ual features on a 0-100% accuracy scale is reported in
Fig. 12. The relative performance of the features gives
vital information to the note designers as to which fea-
ture is performing the best and which feature is more
susceptible to counterfeiting attack.
(a)
(b)
(c)
Fig. 11 Separability of notes in feature space using: (a) ink
features, (b) artwork features, and (c) printing technique
based features.
Fig. 12 Performance of individual features. The features are
along the x-axis and the corresponding accuracies (%) are
along the y-axis.
5 Conclusion
An automatic method for authentication of currency
notes is explored. This research is particularly impor-
tant when the problem of fake bank notes is considered
as a serious problem in many countries. The present ex-
periment considers Indian bank notes as reference. This
study investigates how the security features can be com-
putationally captured in order to automate the authen-
tication process. Exhaustive evaluation of the method
using real life samples brings out the potential of the
approach.
The complexity of the overall system is kept optimal
so that a low cost hardware realization of the proposed
method is feasible. A low cost system is in demand so
that a large scale deployment of such a system becomes
possible. For this purpose, we are in touch with a few
companies who are interested in prototyping such a sys-
tem. Some algorithmic optimization may be needed for
embedded realization of the present system.
Another immediate extension of this study is to
evaluate the method on a dierent test collection. We
are in process of collecting a new set of samples from
another laboratory (dierent from the one from which
we received the current data set) of the Department of
Forensic Sciences. Exploiting new features and method
for authentication is indeed needed to make the system
robust against future counterfeiting eorts. In fact, the
present study does not consider one important secu-
rity feature, namely the watermark feature of the cur-
rency notes. The reason behind this refers to the strange
habits of Indian people scribbling by ink pen over the
blank region on the note where watermark is embedded.
Such scribbling marks make the use of the watermark
feature very sensitive in authenticating bank notes. Our
future eort will explore how to get rid of such scrib-
14 Ankush Roy et al.
bling marks and use the embedded watermark as one
of the security features.
6 Acknowledgement
The authors sincerely thank the questioned document
examiners of the Central Forensic Science Laboratory
(CFSL), Govt. of India for their kind help and cooper-
ation.
References
1. Counterfeit Currency in Canada, Publication of Royal
Canadian Mounted Police, December, 2007
2. R. Kaushal, Fake money circulation boosts black economy
India Today, August 5, 2009.
3. Reserve Bank of India, High Level RBI Group Suggests
Steps to Check Menace of Fake Notes, Press release 2009-
2010/232, Dated 11 August, 2009
4. T. Thompson, Security clampdown on the home PC ban-
knote forgers, The Observer, June, 2004
5. R.L. van Renesse, Paper based document security-a review,
IEEE European Conference on Security and Detection
(ECOS), 75-80, 1997.
6. Steven J. Murdoch, and Ben Laurie, The Convergence
of Anti-Counterfeiting and Computer Security, Lecture at
21st Chaos Communication Congress, Berliner Congress
Center, Berlin, Germany, 2004
7. Department of Financial Services, Ministry of Finance,
Govt. of India, Counterfeit Notes from ATM, Press
release number F No 11/16/2011-FI, 3rd Floor, Jeevan
Deep Building, New Delhi, India, 6th July, 2012.
8. C.D. Seropyan, Means of Preventing Counterfeiting Bank
Notes, US Patent, No. 17,473, January 2, 1857.
9. E. Gotaas, Sensor for verication of genuineness of security
paper, US Patent 5,122,754, June 16, 1992.
10. S.K. Harbaugh, Capacitive verication device for a se-
curity thread embedded within currency paper, US Patent
5,417,316, May 23, 1995.
11. E. Slepyan, A. Kugel and J. Eisenberg, Currency Veri-
cation, US Patent, No. 6,766,045, July 20, 2004.
12. M. Massimo, Device for validating banknotes, EPO Patent,
No. EP 0537513 (A1), April 21, 1993.
13. B.T.Graves, W.J.Jones, D.U.Mennie and F.M.Sculits,
Method and Apparatus for Authenticating and Discriminat-
ing Currency, US Patent, No. 5,960,103, September 28,
1999.
14. F. Takeda and S. Omatu, Recognition System of US Dollars
Using a Neural Network with Random Masks, in Proc. of
the Int. Joint Conf. on Neural Networks, Nagoya, Japan,
Vol. 2, 1993.
15. A. Frosini, M. Gori and P. Priami, A Neural Network-
Based Model for Paper Currency Recognition and Verica-
tion, IEEE Transactions on Neural Networks, 7(6):1482-
1490, 1996.
16. M. Aoba, T. Kikuchi and Y. Takefuji, Euro Banknote
Recognition System Using a Three Layered Perceptron and
RBF Networks, IPSJ Trans. on Mathematical Modelling
and Its Applications, 44:99-109, 2003.
17. K.K. Debnath, S. Ahmed and Md. Shahjahan, A Pa-
per Currency Recognition System Using Negatively Corre-
lated Neural Network Ensemble, Journal of Multimedia,
5(6):560-567, 2010.
18. H. Hassanpour and P.M. Farahabadi, Using Hidden
Markov Models for paper currency recognition, Expert Sys-
tems with Applications, Elsevier, 36:1010510111, 2009.
19. A. Vila, N. Ferrer, J. Mantecon, D. Breton and J.F. Gar-
cia, Development of a fast and non-destructive procedure
for characterizing and distinguishing original and fake euro
notes, Analytica Chimica Acta, 559(2):257-263, 2006.
20. K. Yoshida, M Kamruzzaman, F.A. Jewel and R.F. Sa-
jal, Design and implementation of a machine vision based
but low cost stand-alone system for real time counterfeit
Bangladeshi bank notes detection, in Proc. 10th Int. Conf.
on Computer and Information Technology (ICCIT), pp.
1-5, Dhaka, Dec. 2007.
21. Features & Utility of Paradigm EXC 6700-I, E-Brochure,
Paradigm Cash Systems Pvt. Ltd., 2009.
22. A. Roy, B. Halder and U. Garain, Authentication of Cur-
rency Notes through Printing Technique Verication, in
Proc. of ACM, Indian Conference on Computer Vision,
Graphics and Image processing (ICVGIP), Dec 12-15,
Chennai, India, 2010.
23. R.D. Warner, R.M. Adams and M. Believe, Introduc-
tion to Security Printing, Graphic Arts Center Publishing
Company, ISBN 0883623757, 2005.
24. K.W. Bender, Moneymakers: the secret world of banknote
printing, Wiley-VCH Publishers, 2006.
25. Reserve Bank of India, Detection and Impounding of Coun-
terfeit Notes, Master Circular, released on July 01, 2011.
26. R.C. Gonzalez and R.E. Woods, Digital Image Processing,
Prentice Hall, 2008.
27. R.O. Duda, P.E. Hart and D.G. Strok, Pattern Classica-
tion, Wiley, 2001.
28. J.S. Kelly and B.S. Lindblom (Eds), Scientic Examina-
tion of Questioned Documents, 2nd Edition (Forensic and
Police Science Series), CRC Press, 2006.
29. C. Schulze, M. Schreyer, A. Stahl and T.M. Breuel, Eval-
uation of Gray level Features for Printing Technique Clas-
sication in High Throughput Document Management Sys-
tems, in 2nd Int. Workshop on Computational Forensics
(IWCF), LNCS, Springer Berlin/Heidelberg, 5158:3546,
Washington, DC, USA, August, 2008.
30. C.H. Lampert, L. Mei and T.M. Breuel, Printing Tech-
nique Classication for Document Counterfeit Detection, in
Proc. of Int. Conf. on Computational Intelligence and
Security, pp. 639-644, China, 2006.
31. R.G. Kuehni, Color Space and Its Divisions: Color Order
from Antiquity to the Present, John Wiley and Sons, 2003.
32. K. Franke and S. Rose, Ink-deposition model: The rela-
tion of writing and ink deposition processes, In Proc. of the
Ninth International Workshop on Frontiers in Handwrit-
ing Recognition (IWFHR), pp. 173178, 2004.
33. J. Duchene and S. Leclercq, An optimal transformation
for discriminant and principal component analysis, IEEE
Transaction on Pattern Analysis and Machine Intelli-
gence, 10(6):978-983, 1988.