Review
Pulse coupled neural networks and its applications
M. Monica Subashini
⇑
, Sarat Kumar Sahoo
1
School of Electrical Engineering, VIT University, Vellore, Tamil Nadu, India
article info
Keywords:
PCNN model
Modifications
Applications in image processing
Miscellaneous applications
abstract
This paper surveys the extensive usage of pulse coupled neural networks. The visual cortex system of
mammalians was the backbone for the development of pulse coupled neural network. PCNN (Pulse Cou-
pled Neural Networks) is unique from other techniques due to its synchronous pulsed output, adjustable
threshold and controllable parameters. is Hence the uniqueness of this network utilized in the fields of
image processing. The basic model of PCNN and the consecutive changes implemented, to strengthen
the pulse coupled neural network are discussed initially. Then the applications of PCNN are broadly dis-
cussed. The other miscellaneous applications utilizing pulse coupled neural networks are thrown light in
the last section.
Ó 2013 Elsevier Ltd. All rights reserved.
1. Introduction
Image processing is a common term that covers image segmen-
tation, image registration, image fusion, image thinning, image
enhancement, edge detection, feature extraction, image recogni-
tion, noise removal from image, classification of images, texture
and fabric defects identification, and surveillance. For all these
above mentioned image processing techniques pulse coupled neu-
ral network was found to be a suitable processor. The applications
of PCNN
2
are taken into consideration for a detailed survey. This
network is very much finding its usefulness in diagnosis of diseases
through image processing techniques. A few other applications
other than medicine are also discussed to learn the extensive appli-
cation of pulse coupled neural network. People find less time in
analyzing and interpretating solutions for problems detected.
Hence they seek the help from machines. Machines are thus
trained to perform the functions of a human brain. The human
brain contains about 10 billon neurons and those neurons are the
participants in the parallel information processing system. Artificial
neural networks were developed to bring computers a bit closer to
the brain’s capabilities.
Biological systems have always been an inspiration for develop-
ing algorithms. The mammal’s visual cortex formed the base of
some network models. Cat’s and guinea pig’s visual cortex helped
in developing some digital models. The input information is re-
ceived by the eye. Receptors within the eye (retina) are not sensi-
tive to all the information. The sensitivity is based on color, motion
and intensity. The receptor after receiving the information alters
the behavior of surrounding receptors with respect to the contents
and then forwards to the visual cortex and then the received infor-
mation is analyzed by the brain. The functioning of the visual cor-
tex has to be studied in order to develop algorithms. This is more
complicated than programming of computers. Researchers started
working in the beginning of 1950s (Hodgkin & Huxley, 1952). They
had described the membrane potentials in terms of rate of change
of different chemical elements. In the early 1960s, a mathematical
model was developed by Fitzhugh (1961) based on coupled oscilla-
tors. The dynamics of neurons were in oscillatory process. They de-
scribed the neuron as a two coupled oscillator that are connected
to neighboring neurons. Later in 1990s Arndt, Dicke, Eckhorn,
and Reitboeck (1990) introduced a model on cat’s visual cortex.
In their model, each neuron received input from its own stimulus
and also from the neighboring neurons. The outputs from other
neurons were also an input for the parent neuron. In 1992, Rybak,
Sandler, and Shevtsova (1992) came up with a model based on gui-
nea pigs’visual cortex. The model was similar to Eckhorn except in
equations. Then Johnson and Ritter (1993) implemented the pulse
coupled neural network and suggested a new mechanism with
limited connectivity for information transmission. Also the net-
work was implemented for image processing by eminent research-
ers (Johnson, Kuntimad, & Ranganath, 1995; Johnson & Padgett,
0957-4174/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.eswa.2013.12.027
⇑
Corresponding author. Tel.: +91 416 2202364; fax: +91 416 2243092.
E-mail addresses: monicasubashini.m@vit.ac.in (M. Monica Subashini), sksahoo@
vit.ac.in (S.K. Sahoo).
1
Tel.: +91 416 2202467; fax: +91 416 2243092.
2
Abbreviations used: PCNN, pulse coupled neural networks; ICM, intersecting
cortical model; fMRI, functional magnetic resonance imaging; EM-PCNN, expected
maximization-pulse coupled neural network; BCFCM, bias corrected fuzzy c-means;
IAF, integrate and fire; m-PCNN, multichannel-pulse coupled neural network; LSWT,
lifting stationary wavelet transform; DWT, discrete wavelet transform; LSWT-PCNN,
lifting stationary wavelet transform- pulse coupled neural network; NSCT, non-
sampled contourlet transform; PCNNAI, pulse coupled neural network with aniso-
tropic interconnect ions; FCM, fuzzy-c-means; PCNN-FMI, pulse coupled neural
network-fuzzy mutual information; ADATE, automatic design of algorithms through
evolution; FSD, filter subtract decimate.
Expert Systems with Applications 41 (2014) 3965–3974
Contents lists available at ScienceDirect
Expert Systems with Applications
journal homepage: www.elsevier.com/locate/eswa