Mahmoud Alborzi et al./ Elixir Comp. Sci. & Engg. 41 (2011) 5797-5802 5797
Introduction
Hopfield Network is one of the neural networks which have
been used in the present research. Hopfield network in images
can have useful applications. Fingerprint images which are used
for identification may have noise. For example, it may be
injured. In the present research, we study to what extent
Hopfield Network can improve noisy or defective fingerprint
images and be applied in fingerprint recognition process.
Fingerprint images become binary after recalling to be
calculated for Hopfield network studies. Hamming Distance is
regarded as the best option due to application for binary data.
Hamming Distance will find the closest similar image by
studying differences between each image of input fingerprint
and other images.
Images of fingerprint
Fingerprint of each person is a unique biological trait and
can have effective applications for identification. Identification
system based on biological traits should be distinguished and
can be gathered and stable and bolometric trait of fingerprint has
the mentioned specifications so that it has been regarded as valid
evidence in judicial authorities (Daliri et al, 2007). Images of
fingerprint are fingertips tissue pattern which have unevenness.
Unevenness of the fingertip is due to wrinkles (Jean et al, 1997).
According to analysis and study of wrinkles in fingertip, we can
adjust identify recognition system based on this biological trait.
Preprocessing operations on images of fingerprints images:
In the present research, 40 fingerprints images have been
used as model set for Hopfield Network learning. All images of
fingertips are converted to black and white format and then size
of all images is equal. Primary images of fingertip are in large
dimensions which are difficult to calculate. For this reason,
images will be calculated in smaller sizes. Background of
fingertips is removed and contrast of the images is increased so
that images of fingertips are made more evident. Stages of
preprocessing operations on one of the fingerprint images are
shown in Figure 1. After performing preprocessing on the model
set images, Hopfield operations will be performed.
Figure 1: Preprocessing operations on the fingerprint
images
Recall of fingerprint images
In the present research, all operations on images have been
done in MATLAB software. In this software, the related matrix
will be specified after recall of the images. Matrix elements of
fingerprint images are determined on the basis of gray spectrum
zero to 255 belonging to each one of the image pixels. In the
next stage , images are binary and because the studied method is
Hopfield network and all studied elements in Hopfield network
should be +1 and -1, therefore, all zeros are converted to -1 and
+1s remain unchanged. In the following Figure, matrixes of
fingerprint image are shown. In the first stage, all elements of
matrix are between zero and 255 and are zero and 1 in the next
stage and finally between -1 and +1.
Figure 2: Fingerprint images matrices
Tele:
E-mail addresses: toloie@gmail.com, toloie@srbiau.ac.ir
© 2011 Elixir All rights reserved
Application of hopfield network in improvement of fingerprint recognition
process
Mahmoud Alborzi
1
, Abbas Toloie- Eshlaghy
1
and Dena Bazazian
2
1
Industrial management department, Science and Research Branch, Islamic Azad University, Tehran, Iran
2
Information technology management department, Science and Research Branch, Islamic Azad University, Tehran, Iran.
ABSTRACT
Hopfield Network is able to convert noisy data provided to the network because it can act as
content addressable memory. Hopfield Network can convert images of noisy fingerprint to
the noiseless images or the images with the minimum noise by training through receivable
models set. In this research, fingerprint recognition process has been performed through
Hamming Distance and effect of use of Hopfield network on fingerprint recognition process
has been mentioned. In case those Hamming Distance operations are performed after
Hopfield Networks processing, error of fingerprint recognition will be reduced.
© 2011 Elixir All rights reserved.
ARTICLE INFO
Article history:
Received: 25 September 2011;
Received in revised form:
18 November 2011;
Accepted: 29 November 2011;
Keywords
Fingerprint image,
Noise,
Hopfield Network,
Hamming Distance.
Elixir Comp. Sci. & Engg. 41 (2011) 5797-5802
Computer Science and Engineering
Available online at www.elixirpublishers.com (Elixir International Journal)