30 | P
age
Australian Journal of Information Technology and Communication Volume II Issue I
ISSN 2203-2843
Optimal Image Fusion using Neuro-Fuzzy
Algorithm and SVM
Ms Maninder Kaur1, Ms Pooja2
(Department of CSE,CTIEMT,CT Institution Shahpur,Jalandhar)
1
maninder.khinda43@gmail.com
(Department of CSE, CTIEMT, CT Institution Shahpur,Jalandhar)
2
poojachoudhary80@gmail.com
Abstract: This thesis presents a novel image fusion
method which is suitable for pan-sharpening of
multispectral (MS) bands, based on multi-resolution
analysis. The low-resolution MS bands are sharpened
by injecting high-pass directional details extracted
from the high-resolution panchromatic (Pan) image
by means of the Wavelet and curvelet transform which
is a non-separable MRA whose basis function are
directional edges with progressively increasing
resolution. The fusion of high-spectral but low spatial
resolution multispectral and low-spectral but high
spatial resolution panchromatic satellite images is a
very useful technique in various applications of
remote sensing. Some studies showed that waveletbased image fusion method provides high quality of
the spectral content of the fused image. In this paper
we introduce a new method based on the Wavelet and
curvelet transform using Neuro-Fuzzy which
represents edges better. Since edges play a
fundamental role in image understanding one good
way to enhance spatial resolution is to enhance the
edges. Wavelet and curvelet-based image fusion
method provides richer information in the spatial and
spectral domains simultaneously. It will perform
image fusion using Wavelet and curvelet Transform
with Neuro-Fuzzy Algorithm. This new method has
reached an optimum fusion result. For the
implementation of this proposed work we use the
Image Processing Toolbox under Matlab Software.
Keywords: Image Fusion, Curvelet Transform,
Wavelet Transform, Neuro-Fuzzy, SVM, Peak Signal
to Noise Ratio (PSNR), Root Mean Square Error
(RMSE).
INTRODUCTION
The process of including complementary and redundant
information from different images into one composite
image which includes a better description of the
underlying scene is known as image fusion and these
results in a fused image more useful for human visual
and machine processing. Image fusion strategies are
basically classified into pixel level and region level
approaches.
Pixel level techniques: The set of pixels in the source
image determine each pixel in the fused image.
Basically pixel level techniques are classified into
spatial domain and transform domain techniques.
Region level techniques: In this technique the
segmentation of the images into regions and then based
upon the extracted region fusion is performed.
The process of image fusion c o m b i n e s two or more
images. Different images contain different information
is the main idea behind image fusion. Wavelet
transforms is that in which the transformation should
allow only changes in time extension, but not shape.
This is affected by choosing suitable basis functions
that allow for these Changes in the time. Curvelet are an
appropriate basis for representing images which is
smooth apart from singularities along smooth
curves where the curves have bounded curvature where
objects have a minimum length scale in the image. This
property holds for cartoons and geometrical diagrams
and text.
WAVELET TRANSFORM:
Wavelet transforms is that in which the transformation
should allow only changes in time extension, but not
shape. This is affected by choosing suitable basis
functions that allow for these Changes in the time. A
signal analysis method similar to image pyramids is the
discrete wavelet transform. The main difference is that
while image pyramids lead to an over complete set of
transform coefficients, the wavelet transform results in
a no redundant image. The discrete 2-dim wavelet
transform is computed by the recursive application of
low pass and high pass filters in each direction of the
input image followed by sub sampling. One major
drawback of the wavelet transform when applied to
image fusion is its well known shift dependency this
results in inconsistent fused images when invoked in
image sequence fusion.
A wavelet transform array is synthesized for the product
image and populated from the source images based on a
set of predefined rules. After population, this synthetic
array is inverse wavelet transformed to create the
product image. Graham provides a more detailed
discussion of the application of wavelet theory to image
31 | P
age
Australian Journal of Information Technology and Communication Volume II Issue I
fusion. Our prototype system handles images which are
co registered and the same size, with dimensions which
are powers of two. A wavelet transforms using the
Daubechies basis functions with filter lengths of 4, 12,
or 20 are performed on all input images to be fused.
A significant part of our work cantered on determining
rules to use in combining wavelet transform array
information. The system needed to allow operations to
be performed on individual wavelet array blocks, so that
low and high frequency components could be treated
differently. The system needed to enable the use of
many different combination rules. The approach taken
to achieve this was to identify several primitive
operations that would be required to implement a
variety of combination rules. These primitive operations
act upon individual frequency blocks in wavelet arrays
or on whole wavelet arrays at once. The most useful
primitive operation is to simply take the coefficient with
the maximum amplitude from any input Image at each
location in the wavelet transform array. Another
operation is to average the values in all input wavelet
transform arrays at each location. This operation, if
performed on the entire wavelet array and inverse
transformed, produces a result indistinguishable from
the result of simply averaging the input images. We
found this operator useful when performed on selected
frequency blocks in combination with other operators.
CURVELET TRANSFORM:
The Curvelet transform like the wavelet transform is a
Multiscale transform with frame elements indexed by
scale and location parameters. Unlike the wavelet
transform parameters and the Curvelet pyramid contains
elements with a very high degree of directional specific
city. In addition, the Curvelet transform is based on a
certain anisotropic scaling principle which is quite
deferent from the isotropic scaling of wavelets. The
elements obey a special scaling law where the length of
the support of a frame elements and the width of the
support are linked by the relation width ¼ length2. All
of these properties are very stimulating and have
already leaded to a range of interesting idealized
applications {for example in tomography and in
scientific computation. An understanding of the
Curvelet transform concept opens one's eyes to the fact
that in two and higher dimensions new Multiscale
representations are possible, having properties
unavailable by wavelets and having stimulating
structural features. While it is possible that this new
idea will be quickly forgotten with the passage of time
we feel that the very novel features of the transform anisotropy, anisotropy scaling - compel further
investigation for the moment.
The Curvelet transform (CVT) is a multi-scale
transform proposed by Candes and Donoho and is
derived from the Ridgelet transform The Curvelet
transform is suited for objects which are smooth away
from discontinuities across curves then the Fourier
ISSN 2203-2843
Transform does not handle point's discontinuities well
because a discontinuity point affects all the Fourier
Coefficients in the domain. Moreover, Wavelet
transform handles point discontinuities well and doesn't
handle curve discontinuities well. Curvelet transform
handles curve discontinuities well as they are designed
to handle curves using only a small number of
coefficients and Curvelet transform has several
applications in various areas such as image denoising,
image fusion, Seismic exploration, Turbulence analysis
in fluid mechanics and so on. Curvelet Transformation
is an enhancement technique to reduce image noise and
to increase the contrast of structures of interest in
image. Compared to other techniques, this method can
manage the vagueness and ambiguity in many image
reconstruction applications efficiently.
NEURO-FUZZY (ANFIS):
The Neuro-fuzzy term was born by the fusing of these
two techniques Neuro and fuzzy. As each researcher
combines these two tools in different way, then, some
confusion was created on the exact meaning of this
term. Still there is no absolute consensus but in general,
the Neuro-fuzzy term means a type of system
characterized for a similar structure of a fuzzy
controller where the fuzzy sets and rules are adjusted
using neural networks tuning techniques in an iterative
way with data vectors (input and output system data). A
Neuro-fuzzy technique called Adaptive network based
fuzzy inference system (ANFIS) has been used as a
prime tool. Adaptive network based fuzzy inference
system (ANFIS) is a Neuro fuzzy technique where the
fusion is made between the fuzzy inference system and
the neural network. In ANFIS the parameters can be
estimated in such a way that both the Sugeno and
Tsukamoto fuzzy models are represented by the ANFIS
architecture. Again with minor constraints the ANFIS
model resembles the Radial basis function network
(RBFN) functionally.
This ANFIS methodology comprises of a hybrid system
of fuzzy logic and neural network technique. The fuzzy
logic takes into account the imprecision and uncertainty
of the system that is being modelled [7]. Using this
hybrid method, at first an initial fuzzy model along with
its input variables are derived with the help of the rules
extracted from the input output data of the system. Next
the neural network is used to fine tune the rules of the
initial fuzzy model to produce the final ANFIS model of
the system. In this proposed work ANFIS is used as the
backbone for the identification of real world systems.
SUPPORT VECTOR MACHINE:
It is primarily a classifier in which Width of the margin
between the classes is the optimization criterion, i.e.
empty area around the decision boundary defined by the
distance to the nearest training patterns. These are
called support vectors. The support vectors change the
prototypes with the main difference between SVM and
traditional template matching techniques is that they
characterize the classes by a decision boundary which is
32 | P
age
Australian Journal of Information Technology and Communication Volume II Issue I
not just defined by the minimum distance function. The
concept of (SVM) Support Vector Machine was
introduced by Vapnik. The objective of any machine
that is capable of learning is to achieve good
generalization performance, given a finite amount of
training data. The support vector machines have proved
to achieve good generalization performance with no
prior knowledge of the data.
The principle of an SVM is to map the input data onto a
higher dimensional feature space nonlinearly related to
the input space and determine a separating hyper plane
with maximum margin between the two classes in the
feature space. The SVM is a maximal margin hyper
plane in feature space built by using a kernel function.
These results in a nonlinear boundary in the input space
and the optimal separating hyper plane can be
determined without any computations in the higher
dimensional feature space by using kernel functions in
the input space. There are some commonly used kernels
include:a) Linear Kernel
ISSN 2203-2843
The PSNR block computes the peak signal-to-noise
ratio, in decibels, between two images. PSNR
represents a measure of the peak error. PSNR is the
ratio between the maximum possible power of a signal
and the power of corrupting noise that affects the
fidelity of its representation and the PSNR of the fusion
result is defined as follows:
PSNR =
[
/
]
(2)
Where, R is the maximum fluctuation in the input
image data type. Higher the value of the PSNR, better
the performance of the fusion algorithm.
PROPOSED WORK: Optimal Fusion Method
The proposed work of image fusion is shown in the
following figure:-
K(x, y) = x, y
b) Polynomial Kernel
K(x, y) = (x. y+1) d
SVM Algorithm
i.
ii.
iii.
Define an optimal hyper plane.
Extend the above definition for non linear
separable problems.
Map data to high dimensional space where it is
easier to classify with linear decision surfaces.
PRINCIPAL COMPONENT ANALYSIS (PCA)
PCA used for reducing the multidimensional data sets
to lower dimensions PCA algorithm for the fusion of
image is calculated the Eigen values and the Eigen
vectors [17].
MEAN SQUARE ERROR (MSE)
Mean Square Error (MSE) a commonly used reference
based assessment metric is the Mean Square Error
(MSE). The MSE represents the cumulative squared
error between the reconstructed image and the original
image is given by the following equation:
MSE = ∑
[ (
, )
(
, )]
∗
(1)
Where, M and N are the number of rows and columns
in the input images, respectively. Lower the value of
MSE the lower the error.
PEAK SIGNAL TO NOISE RATIO (PSNR)
Figure 1 Image Fusion using Neuro-Fuzzy Algorithm & SVM
33 | P
age
Australian Journal of Information Technology and Communication Volume II Issue I
ISSN 2203-2843
RESULTS AND DISCUSSION
In the following figures, result of proposed algorithm is
highlighted. The comparison is made with the existing
method. The existing method is based on Waveletcurvelet with PCA while in our methodology we have
optimised the results by applying ANFIS and SVM.
Figure 6 Graphical representation of RMSE Comparison
Figure 2
CT Train Image
Figure 7 Comparison of PSNR
Figure 3 MR Train Image
Figure 5 Comparison of RMSE
Figure 8 Graphical representation of PSNR Comparison
CONCLUSION
The paper has presented a new trend in the fusion of
digital image, MRI and CT images which are based on
the Wavelet and Curvelet transform. A comparison
study has been made between the traditional wavelet
fusion algorithm and the proposed Curvelet fusion
algorithm. The experimental study shows that the
application of the Curvelet transform in the fusion of
MR and CT images is superior to the application of the
traditional wavelet transform. The obtained Curvelet
fusion results have minimum MSE and PSNR than in
wavelet fusion results. At last, these fusion methods are
used to detect brain tumors. In vision, the Neuro-fuzzy
algorithm is proposed for image fusion in this paper
34 | P
age
Australian Journal of Information Technology and Communication Volume II Issue I
acquires better fusion result. In objective evaluation
criteria, Curvelet fusion characteristic are superior to
wavelet transform.
REFERENCES
[1] Vishal P.Tank, Divyang D. Shah,Tanmay V. Vyas,
Sandip B. Chotaliya Manthan S. Manavadaria in 2013.
They are purpose Image Fusion Based on Wavelet and
Wavelet and curvelet Transform.
[2] Jianwei Ma and Gerlind Plonka in 2012. He purpose
A Review of Curvelets and Recent Applications.
Multiresolution methods are deeply related to image
processing, biological and computer vision, scientific
computing, etc.
ISSN 2203-2843
[9] J. L. Starck, E. J. Cand‘es and D. L. Donoho, “The
curvelet transform for image denosing,” IEEE Trans.
Image Processing, vol. 11, 2002, pp. 670-684.
[10] J. L. Starck, E, J. Cand`es, and D. L. Donoho,
“Gray and Color Image Contrast Enhancement by the
Curvelet Transform,” IEEE Trans. Image Processing,
vol. 12, no. 6, 2003, pp. 706-717.
[11] E. J. Cand‘es, “Ridgelets: Theory and
Applications,” Ph.D. Thesis, Department od Statistics,
Stanford University, Standford, CA, 1998.
[12] D. L. Donoho, “Digital ridgelet transform via
rectopolar coordinate transform,” Stanford Univ.,
Stanford, CA, Tech. Rep, 1998.
[3] Smt.G. Mamatha (Phd), L.Gayatri in 2012. They are
purposed AN IMAGE FUSION USING WAVELET
AND CURVELET TRANSFORMS.
[13] D. L. Donoho, “Orthonormal ridgelets and linear
singularities,” SIAM J. Math Anal., vol. 31, no. 5, 2003,
pp. 1062-1099.
[4] T. Ranchin and L. Wald, “Fusion of High Spatial
and Spectral Resolution images: The ARSIS Concept
and Its Implementation,” Photogrammetric Engineering
and Remote Sensing, vol. 66, 2000, pp. 49-61.
[14] M. I. Smith, J. P. Heather, "Fusion Technology
Review of Image in 2005," Proceedings of the SPIE,
Volume 5782, pp. 29-45, 2005.
[5] L. Wald, T. Ranchin and M. Mangolini, “Fusion of
Satellite images of different spatial resolution:
Assessing the quality of resulting images,”
Photogrammetric Engineering and Remote Sensing, vol.
63, no. 6, 1997, pp. 691-699.
[6] J. Nunez, X. Otazu, O. Fors, A. Prades, V. Pala and
R. Arbiol, “Multiresolution-based image fusion with
addtive wavelet decomposion,” IEEE Transactions on
Geoscience and Remote sensing, vol. 37, no. 3, 1999,
pp. 1204-1211.
[7] E. J. Cand‘es, “Harmonic analysis of neural
networks,” A ppl. Comput.Harmon. Anal., vol. 6, 1999,
pp. 197-218.
[8] E. J. Cand`es and D. L. Donoho, “Curvelets- A
surprisingly effective non adaptive representation for
objects with edges,” in Curve and Surface Fitting:
Saint-Malo, A. Cohen, C.Rabut, and L.L.Schumaker,
Eds. Nashville, TN: Vanderbilt Univ. ersity Press, 1999.
[15] Yong Yang, "Multi modal Medical Image Fusion
through a New DWT Based Technique", 4th
International Conference on Bioinformatics and
Biomedical Engineering, pp 1-4, 2010.
[16]
Chandrakanth.R,
SaibabaJ,Varadan.G,
Raj.PA,"Fusion of High Resolution Satellite SAR and
Optical Images "International Workshop on MultiPlatform/Multi-Sensor Remote Sensing and Mapping ,
pp 1-6, Jan 2011.
[17] Anand, T. S., K. Narasimhan, and P. Saravanan.
"Performance evaluation of image fusion using the
Multi-Wavelet and Curvelet transforms." Advances in
Engineering, Science and Management (ICAESM),
2012 International Conference on. IEEE, 2012.