Banknote Simulator for Aging and Soiling Banknotes using Gaussian
Models and Perlin Noise
Sangwook Baek
1
, Sanghun Lee
1
, Euison Choi
2
, Yoonkil Baek
2
and Chulhee Lee
1
1
School of Electrical and Electronic Engineering, Yonsei University, Seoul, Republic of Korea
2
Advanced Development, R&D Nautilus Hyosung Inc., Seoul, Republic of Korea
Keywords: Banknote Simulator, Soling Banknotes, Gaussian Model, Perlin Noise.
Abstract: In this paper, we propose a banknote simulator that generates aged and soiled banknotes. By analyzing the
characteristics of circulating banknotes, we developed Gaussian brightness models for gray level changes of
circulating banknotes. In addition, the Perlin noise model was used to simulate soiling. The proposed
algorithm was tested using US Dollars (USD) and the experimental results show that the proposed method
effectively simulated soiled banknote images from new banknote images.
1 INTRODUCTION
As financial transactions greatly increase, a large
number of cash transactions are conducted through
automated systems. Financial automation systems
are widely used in many applications such as
automated teller machines (ATM). These systems
perform a number of functions, which include
banknote classification, fake banknote detection, etc.
On the other hand, as the circulation period of
banknotes increases, the recognition accuracy of
aged banknotes decreases. In many cases, ATM
software (SW) needs to be upgraded to deal with
aged banknotes. However, when new banknotes are
introduced, ATM developers also need to develop
classification programs based on the new banknotes
and all the parameter determinations for ATM SW.
Consequently, as more aged banknotes start
circulating after several months, ATMs make higher
errors in terms of banknote classification and
validation. In general, updating the ATM SW is
time-consuming and expensive. Therefore, there is a
great need for good simulators that can produce
street quality banknotes from new banknotes. Some
aging effects include soiling, creasing, and edge
blurness.
Several authors have studied aged banknotes that
were generated using mechanical and chemical
circulation simulators. However, there has been little
research into simulated circulating banknotes that
use image processing techniques.
In this paper, after analyzing the characteristics
of aged banknotes, we propose Gaussian brightness
models that can be used for banknote aging. Next,
we combine the Gaussian brightness models with the
Perlin noise model to simulate aged banknotes.
The rest of this paper is organized as follows:
Section 2 describes the Gaussian brightness models.
Section 3 presents a description of aged banknote
synthesis using the Gaussian models and the Perlin
noise model, and concluding remarks are drawn in
Section 4.
2 GAUSSIAN BRIGHTNESS
MODELS
2.1 Datasets
We acquired banknote images of real circulating
banknotes with a contact image sensor (CIS). The
image resolution was 50 dots per inch (DPI). We
used USD 1 banknotes for data analyses and
parameter estimations. The total number of
banknotes was 2000.
2.2 Observation and Measurement
As the circulation period of banknotes increases,
bright areas of banknotes generally become darker,
and dark areas become brighter (see Figure 1).
Based on this observation, we measured the aging
Baek, S., Lee, S., Choi, E., Baek, Y. and Lee, C.
Banknote Simulator for Aging and Soiling Banknotes using Gaussian Models and Perlin Noise.
DOI: 10.5220/0006112402890292
In Proceedings of the 6th International Conference on Pattern Recognition Applications and Methods (ICPRAM 2017), pages 289-292
ISBN: 978-989-758-222-6
Copyright
c
 2017 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved
289

level of banknotes using Otsu’s threshold. This
method separates a banknote image into two groups
and calculates the average difference of the two
groups. When the difference of the banknote is
small, it can be said that the banknote has a low
level of soiling. Using the average differences, we
classified the dataset into three groups: group A
(relatively new), group B (medium aging), and
group C (heavily aged).
Figure 1: Examples of banknote image (a) new banknote,
(b) soiled banknote.
2.3 Gaussian Models for Banknote
Aging
First, local registration was applied for the banknote
image dataset to a reference banknote (new). After
registration, we measured the brightness changes
within the dataset.
Figure 2 shows some examples of the brightness
distributions of the dataset for three areas whose
brightness levels in the reference image range from a
bright area, a medium area and a dark area.
Although there are some variations, it can be seen
that the bright area (more than 128) became darker,
while the dark area (less than 80) became slightly
brighter. The brightness of the middle area (96-112)
showed few changes. We divided the brightness into
13 intervals and modeled each interval using a
Gaussian distribution. We repeated this procedure
for each of the three groups: group A (relatively
new), group B (medium aging), and group C
(heavily aged). The parameters (means and
variances) of the Gaussian models were computed
from the training dataset (see Figure 3).
Figure 2: Example of brightness change in circulated
banknote.
Figure 3: Histograms of each class (a) dark area, (b)
middle area, (c) bright area.
ICPRAM 2017 - 6th International Conference on Pattern Recognition Applications and Methods
290

3 SYNTHESIS OF BANKNOTE
IMAGES
3.1 Synthesis of Aged Banknotes
Using the Gaussian models, we simulated aged
banknotes from new or nearly new banknotes. First,
we chose the group (A, B or C) and then determined
the degree of aging by randomly selecting the
variations of the Gaussian distributions.
We also preserved the correlations between the
neighboring pixels of the simulated banknotes.
Figure 4 shows an example of an artificially soiled
banknote image using the proposed Gaussian
brightness modeling.
Figure 4: Examples of banknote images (a) real banknote,
(b) soiled banknote using Gaussian model.
3.2 Perlin Noise
Figure 5: Examples of banknote images (a) Perlin noise
map, (b) soiled banknote using Perlin noise.
In order to simulate the natural look of real aged
banknotes, the Perlin noise model was used. Perlin
noise is a sort of gradient noise, which is rescaled
and added into itself to create fractal noise. In the
proposed method, the Perlin noise model (eight
times scaled synthesis) was used to achieve a natural
aging look (see Figure 5). Figure 6 shows examples
of generating Perlin noise.
Since brightness changes need to be different
depending on the target pixel brightness, Perlin
noise was partially applied. In bright areas, Perlin
noise was subtracted while it was added in dark
areas.
Figure 6: Examples of Perlin noise.
4 CONCLUSIONS
In this paper, we proposed an algorithm to generate
artificially aged banknote images. The proposed
method is based on the Gaussian brightness model
and Perlin noise. The proposed simulator can be
used to generate aged banknotes from new
Banknote Simulator for Aging and Soiling Banknotes using Gaussian Models and Perlin Noise
291

banknotes, which can be used to develop robust
ATM SW.
REFERENCES
Kyrychok, T., Shevchuk, A., Nesterenko, V., Kyrychok, P.,
2013. Banknote paper deterioration factors: circulation
simulator method. In BioResources, 9:710-724.
Bartz, W. J., Crane, T. T., 2006. Circulation simulator
method for evaluating bank note and optical feature
durability. In Proc. of SPIE-IS&T Electronic Imaging,
6075:607505.
Perlin, K., 1985. An Image Synthesizer. In ACM Siggraph
Computer Graphics, 19:287-296.
Otsu, N., 1979. A threshold selection method from gray-
level histograms. In IEEE Trans. Syst., Man, Cybern.,
SMC-9:62-66.
ICPRAM 2017 - 6th International Conference on Pattern Recognition Applications and Methods
292