International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)

Web Site: www.ijettcs.org Email: [email protected]

Volume 3, Issue 5, September-October 2014

ISSN 2278-6856

PERFORMANCE EVALUATION OF
INTEGRATED DHE, PCA & MAX-DCT
BASED DIGITAL IMAGE FUSION
Shivdeep kaur1, Dr. Rajiv Mahajan2
Dept of Computer Science & Engineering, GIMET

Abstract
In this paper, a survey on the different image fusion
techniques is presented. Image fusion is a method to
combine the related information from a set of images into
a single image where the resultant fused image will be
having more information in image. It is found from the
survey that vast majority of current researchers have
ignored portion of the popular issues of vision processing
like image de-noising, image enhancement, and image
restoration. So to overcome these problems, we also
devised a new algorithm in this paper. This algorithm
introduces an efficient approach for fusion of multi-focus
images based on variance evaluated in discrete cosine
transform domain. This paper has proposed another
method which will integrate the PCA, Max-DCT and
dynamic histogram equalization to improve the outcome
further. We have implemented the proposed algorithm in
MATLAB using image processing toolbox. From the
experimental results, it is proved that the proposed
algorithm outperforms over the accessible methods.
Index Terms:- Image fusion, Multi-focus, Visual Sensor,
DCT, and PCA.

1. Introduction
Image fusion [1] is an important subject in image
processing. Image fusion is a system of combining the
applicable data from a set of images into a solitary image
where the resultant fused image will be more instructive
image. With fast advancement in technology, it is currently
conceivable to achieve data from multi-source images to
make an excellent fused image. The outcome of image
fusion is alternate image that keeps the most desirable data
and characteristics of input Image. Image fusion is a
process in which images are obtained from distinctive
sensors by a specific algorithm so that the resultant image
is more consistent, clear, and reasonable. Image fusion
method like discrete cosine transform is more appropriate
and efficient in continuous framework. A efficient method
for fusion of multi focus images is based on variance
calculated in DCT domain. In all sensor networks [2],
every sensor can attain, make and exchange information
from one to another. Visual Sensor Networks (VSN) refers
to a framework with a set of cameras that are used to
enlarge the resources and supervising of numerous points.
In visual sensor network, sensors are cameras which can
record either video sequence or pictures. A well-known
feature of visual sensor is to generate great amount of data.
This feature of visual sensor conveys just the supportive
data which is symbolized at a conceptualized stage. The

Volume 3, Issue 5, September-October 2014

objective of image fusion, is to decrease the amount of

information during network transmissions, is to produce
another image that is more proper for further imageprocessing tasks. The images used in image fusion have to
be registered. Pixel level fusion technique is used to
increase the spatial resolution of the multispectral image.

2. Adaptive Histogram Equalization

Adaptive histogram equalization is a computer image
processing strategy used to improve contrast in images.
Adaptive histogram equalization is an outstanding contrast
enhancement for both normal images and medicinal
images and other at first non visual images It varies from
common histogram equalization in the respect that the
adaptive techniques computes few histograms, each one
comparing to a distinct segment of the image, and uses
them to redistribute lightness estimation of the image. In
image fusion process, fusion methodology may corrupt the
sharpness of the fused image so to overcome this issue of
poor brightness adaptive histogram equalization will be
used to improve the results further. We can say that
adaptive histogram equalization will come in action to
protect the brightness of the fused image.
Need of adaptive histogram equalization: Convert the
reference image to cosine transform. At that point, inverse
cosine transform is applied to reference image. When
inverse cosine transform is applied then noise is occurred
in fused image then adaptive histogram is applied to
remove noise and color artifacts which will introduced due
to transform domain method i.e. DCT
3. Levels of Image Fusion
The primary objective of image fusion is to generate a
fused image that gives the complete and consistent data.
Image fusion perform at three separate levels i.e. pixel,
feature and decision, expects to accomplish the more
correct, complete and consistent image description of the
same scene.

Fig 1: Three levels of image fusion

Page 1

International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)

Web Site: www.ijettcs.org Email: [email protected]
Volume 3, Issue 5, September-October 2014
3.1 Pixel Level Fusion
In pixel level image fusion [3], firstly registration of
images happens. At that point the pre-processing is
preformed. Pixel level fusion meets expectations straight
on pixels of source images. In pixel level image fusion,
various requirements are focused on the fused results; the
fusion methodology should protect all important data in
the source image, the fusion should not present any
component. Pixel level fusion turns into the main principle
since it can secure unique data of source image however
much as could be expected.
3.2 Feature Level Fusion
Feature level fusion procedure is the second phase of
processing where image fusion may happen. Fusion at the
feature level needs removal of features from the input
images. Features could be pixel intensities or edge and
composition characteristics. The Various kinds of features
are measured depending on the way of images and the
application of the fused image. The features includes the
removal of feature primitives like edges, area, shape, size,
length or image fragments, and features with comparable
intensity in the images to be fused from different types of
images of the comparative geographic range. These
features are then merged with the related features
introduce in the other data images through a predetermined choice procedure to form the final fused image.
Decision level fusion joins the results from various
algorithms to yield a final fused decision.

Fig 3: Feature Level Image Fusion [4]

3.4 Decision Level Fusion
Decision level fusion is a very high level of fusion which
joins the results from various algorithms to make a final
fused decision.

4. Literature Review
R. Amutha et al. (2014) Discrete Cosine Transform
based fusion of multi-focus images for visual sensor
networks [5] has discussed that simple and efficient
multi-focus image fusion framework clearly planned for
wireless visual sensor framework prepared with resource
constrained, unsafe setting like battlefields . The fusion of
multi-focus images is focused around higher esteemed
Alternating Current coefficients computed in Discrete
Cosine Transform domain. Discrete cosine transform
defeats the computation and energy confinement of low
power gadgets and is explored as far as image quality and
computation energy. It confirm the significant efficiency
enhancement of the proposed system in yield quality and

Volume 3, Issue 5, September-October 2014

ISSN 2278-6856

energy consumption, when contrasted with other fusion

techniques in DCT domain Vivek Kumar gupta et al.
(2013) Analysis of Image Fusion techniques over
Multispectral and Microwave SAR images [6] has
examined that in remote sensing program the raising
availability of space endured sensors give inspiration to
diverse image fusion algorithms. Remote sensing image
fusion plan at coordinating the data transfer by
information picked up which conceal distinctive parts of
the electromagnetic spectrum at diverse spatial, temporal
and spectral determination; we can secure multi-temporal,
multi-resolution and multi-frequency image data for
purpose of feature extraction, modeling and classification.
The fused picture is much serviceable for human
information. Fused image is more helpful for automatic
computer analysis task for instance feature extraction,
segmentation and object recognition. Om parkash et al.
(2013) Biothogonal wavelet transform based image fusion
using absolute maximum fusion rule [7] has analyzed that
the goal of image fusion is to bring appropriate data out of
two or more images of the same area into a solitary image
which is much informatory and is more suitable for human
information. Spatial domain based operations make spatial
distortions in the fused image. Spatial domain distortion
might be totally overseen by the use of wavelet transform
based image fusion processes. Using supreme greatest
fusion rule wavelet coefficients at unique decomposing
levels are fused. Two weighty characteristics wavelet
symmetry and linear phase of BWT have been took
preferences (exploited) for image fusion in light of the fact
that they can ensure edge data. It has been demonstrated
that the wavelet transform technique improve fusion
quality by decreasing loss of vital data usable in solitary
images. Mohammed Hossny et al. (2013) Image Fusion
Metrics: Evolution in a Nutshell [8] has discussed that
image fusion methodology join together various images
into an individual enlightening image. Image fusion
metrics have created from image processing variance
metrics. In Image fusion metrics: evolution in a nutshell
the evolution of objective image fusion performance
metrics and their subjective and goal acceptance. It clarify
in what way fusion execution metric create starting with
image difference measurement, its understanding into
image fusion connections, it limit weighting component
and the acceptance operation. Desale, R.P. et al. (2013)
Study and analysis of PCA, DCT & DWT based image
fusion techniques. [9] has examined the different image
fusion methods such as PCA (principal Component
Analysis), DCT (Discrete Cosine Transform)and DWT
(Discrete
Wavelet
Transform)
based image fusion
methods. Authors have recommended the selection of
DWT based fusion method for high quality and exactness
applications. In this paper two algorithms focused around
DWT are proposed for example, Pixel averaging and
maximum pixel replacement algorithm. The execution of
above said DWTs have been contrasted and the PCA and
DCT fusion techniques. The examination has been carried
out focused around seven parameters named as PSNR,
MSE, Normalize absolute error, Maximum Difference,
Page 2

International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)

Web Site: www.ijettcs.org Email: [email protected]
Volume 3, Issue 5, September-October 2014
Average difference, Normalized Cross-Correlation and
structural content. The results portrays that the execution
of DWT based fusion strategies is altogether better as
contrast with an alternate routines for image fusion. Kiran
parmar et al. (2012) A Comparative Analysis of
Multimodality Medical Image Fusion Methods [10] has
analyzed that the outline is to improve the image content
by fusing images like computer tomography and magnetic
resonance imaging images; magnetic resonance imaging
gives high quality data on delicate tissue while computed
tomography gives better information about substantial
tissue. Fusing these two sorts of images create a complex
image which is more instructive than different signals gave
by an individual modality. Image fusion has turned into a
regular operation used within medical diagnostics and
therapy. Fast Discrete Curve let Transform using Wrapper
algorithm based image fusion method, has been execute,
examine and contrasted with Wavelet based Fusion
Technique. Fusion of images concentrates at diverse
purposes, power and by distinctive systems helps doctor to
withdraw the qualities that may not be normally visible in
an individual image by different modalities. Rahul K Kher
et al. (2012) Analysis of CT and MRI Image Fusion using
Wavelet Transform [11] has analyzed that the
arrangement is to redesign the image content by fusing
images like computer tomography and magnetic resonance
imaging images, in order to supply more data to the doctor
and clinical therapy planning system. The objective is to
exhibit the application of wavelet transformation to
multimodality medical image fusion. This work secure the
choice of wavelet function, the use of wavelet based fusion
algorithms on medical image fusion of computer
tomography and medical resonance imaging, execution of
fusion regulation and the fusion image standard
assessment. The fusion execution is focused around the
root mean square error and peak signal to noise ratio.
Wang Xin et al. (2011) A New Method for Multi-Focus
Image Fusion Using Countour let Transform [12] has
studied that image fusion focused around Contour let
transform will make a pseudo-Gibbs impact for the
nonattendance of interpretation invariance. We proposed a
system which combine Contour let transform with image
blocking fusion to clarify this matter. There is another
fusion rule focused around Contour let transform which
was offer to get the starting fused image. Source images
and the introductory fused image were divided into
indistinguishable size image blocks. The source image
pieces, which were much near to the relating starting fused
image pieces, were chosen as the closing fused image
blocks. For the frontier fragment between the agreeable
portion and blurring one, the starting fusion results were
kept. It has been demonstrate that this technique can expel
the image distortion appear from Contour let transform
deficiency of interpretation invariance. The fusion effect is
powerful than the impact of image blocking. Ujwala Patil
et al. (2011) Image fusion using hierarchical PCA [13]
proposed that combining two or more registered images of
the indistinguishable place to get numerous educational
images is called image fusion. Principal component

Volume 3, Issue 5, September-October 2014

ISSN 2278-6856

analysis is a prominent approach for feature extraction and

dimension reduction. Image fusion algorithm combines
pyramid and principal component analysis techniques and
take away the quality analysis of hierarchical principal
component analysis fusion algorithm without implication
image. There is a developing requirement for the quality
examination of the fusion algorithms. We show fusion
using wavelet and principal component analysis fusion
methods and takeout creation analysis for these fusion
strategies using unique quality measures for variety of
information sets and demonstrate that proposes image
fusion using hierarchical principal component analysis is
useful for the fusion of multimodal imaged. Dr.M.N.Giri
Prasad et al. (2011) A novel approach of image fusion on
MR and CT images using wavelet transform [14] has
analyzed that almost all areas of medical detection are
affected by the digital image processing. At the point when
an image is transformed for visual clarification, the human
eye is chosen how successfully a particular technique meets
expectations. For medical recognizable proof, Computed
Tomography issues the chief data on denser tissue.
Magnetic Resonance Image issues great data on delicate
tissue. Wavelet transform fusion is characterized by
considering the wavelet transforms of the ensured
information images mutually with the fusion rule. The
inverse wavelet transform is figured, and the fused image
is recreated. Despite the fact that wavelets change some
ordinary properties, every wavelet has a different image
decompression and remarking qualities that prompt
distinctive fusion results. However if a wavelet transform
and a conventional transform like Principal Component
Analysis transform is incorporated, better fusion outcome
may be accomplished
Ahmed Abd-el-kader et al. (2011) Performance Measures
for Image Fusion Based on Wavelet Transform and Curve
let Transform [15] has discussed that curve let transform
is a recently-build multi-scale transforms, which is further
acceptable for items with curves. Image fusion implies the
fusing images into an image that has the great data without
making features that are narrative in the specific images.
Two well-known applications of image fusion are
discovered; fusion of multi-focus images and fusion of
multi exposure images. Fusion results were assessed and
adjusted as indicated by three measures of execution; the
entropy (H), the mutual information (MI) and the amount
of edge information (QABIF). The three execution
measures have demonstrated that the curve let based image
fusion algorithm gives better fused image than the wavelet
algorithm. The fused image has a prevalent eye
observation than the data ones. Xing Su-xia et al. (2011)
Image Fusion Method Based on NSCT and Robustness
Analysis [16] has proposed that infrared and visible
image fusion methods can raise the image contrast, and
expand the night vision effective. The repeated infrared
and visible images from the comparative scene were
unclear by non-sub sampled contour let transform; follow
the
estimated
mass
averaged,
high-frequency
characteristics component as per the weighted of the
regional standard deviation portion; then the fusion image
Page 3

International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)

Web Site: www.ijettcs.org Email: [email protected]
Volume 3, Issue 5, September-October 2014
is acquired by reverse non-sub sampled contour let
transform; the fusion images were contrast with the result
obtained by Laplace transform, wavelet transform and
contour let transform during a vast numeral of trails, and
the strength examination was done through the noise
experiment. Non-sub sampled contour let transform can
achieve prevalent fusion result, and high quality. Ghimire,
D et al. (2011) Nonlinear Transfer Function-Based Local
Approach for Color Image Enhancement. [17] has
proposed a technique for image enhancement in HSV
space focused around the local processing of image. In this
technique an upgrade is connected just on V component
and H and S component are not changed during
improvement so that the first color of the improved image
is not modified. Based upon the subjective and objective
execution assessment, the proposed system has
demonstrated proficient in image improvement. The
destination criteria like Detail Variance, Background
Variance and statistical attributes demonstrates that
proposed strategy deliver better images in comparison to
other systems like histogram equalization and AINDANE.

5. Gaps in Literature Survey

The related work on image fusion algorithms has shown
that the most of existing algorithms are time consuming in
nature and also degrades the brightness of the fused
images. The main reason behind this is that the most of
researchers have neglected one of the following or both:
1. No appropriate image enhancement technique is used
to overcome the poor brightness of the fused image.
2. No transform domain method is used to speed up the
fusion algorithms.
3. No method is considered for complex background
images so it becomes difficult to fused images with
complex background.

6. Problem Definition
The main objective of image fusion is to combine
information from multiple images of the same scene in
order to deliver only the useful information. The discrete
cosine transforms based methods of image fusion are more
suitable and time-saving in real-time systems using DCT
based standards of still images. In this dissertation an
efficient approach for fusion of multi-focus images based
on variance calculated in DCT domain is presented. This
research work propose a new technique which will
integrate the higher valued Alternating Current
coefficients calculated in Discrete Cosine Transform
domain based fusion with principle component analysis
and adaptive histogram equalization to reduce the color
artifacts which will be introduced due to the transform
domain method i.e. DCT. The fusion process may degrade
the sharpness of the fused images so to overcome this
problem adaptive histogram equalization will be used to
enhance the results further. The use of PCA is for
increasing the speed of the AC-DCT based fusion method;
because in color images AC-DCT will be applied on each
color separately which is time consuming in nature. The
performance of image fusion is usually evaluated in terms
of accuracy, PSNR and speed etc.

Volume 3, Issue 5, September-October 2014

ISSN 2278-6856

7. Research Methodology
To attain the objective, step-by-step methodology is used in
this dissertation. Subsequent are the different steps which
are used to accomplish this work. Following are the
various steps used to accomplish the objectives of the
dissertation.
The steps are as follows:
Proposed algorithm
Step 1: Input images: Input 2 images image 1 and image 2 in which image1 is
left blurred and image 2 is right blurred.
Step 2: RGB2PCA: Now RGB to PCA conversion will be done based upon the
certain vector values. Also division of each PCA image will also be done into
PCA1, PCA2 and PCA3.
Step3: Apply AC-DCT based fusion: The next step is to apply AC-DCT based
fusion on first PCA as highest variations found on the first PCA plane. And
chrominance fusion will come in action for other PCA planes i.e. PCA2 and
PCA3.
Step4: Concatenation: Now concatenate the result of each plane and get the
fused image.
Step5: Dynamic histogram equalization: Now dynamic histogram equalization
will come in action to preserve the brightness of the fused image.

8. Experimental Set-up
In order to execute the proposed algorithm, design and
implementation has been carried out in MATLAB using
image processing toolbox. Keeping in mind the end goal to
do cross validation we have likewise executed the
upgraded DCT based image fusion using nonlinear
enhancement. The created methodology is analyzed
against some well-known image fusion techniques
accessible in literature. After these comparisons, we are
looking at proposed methodology against DCT using some
execution measurement. Result demonstrates that our
proposed methodology gives superior results than the
current techniques.
8.1 Experimental Results
Fig.4 (a) is demonstrating the left blurred image and fig.4
(b) is demonstrating the right blurred image. The general
target is to join significant data from various images into a
solitary image that is more enlightening and suitable for
both visual perception and further computer processing.

Fig. 4(a): Left blurred image

Fig. 4(b): Right blurred image

Page 4

International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)

Web Site: www.ijettcs.org Email: [email protected]
Volume 3, Issue 5, September-October 2014

Fig. 5 Max-DCT based image fusion

Figure 5 has demonstrated the output image taken by MaxDCT. The output image has contained too much brightness
and color unevenness as contrast with original blurred
images to be fused.

Fig.6 Final proposed image

Figure 6 has demonstrated the output image taken by the
proposed image fusion with dynamic histogram stretching.
The image has contained the adjusted color and brightness
as the original images to be fused. The nature of output
image is great with our proposed strategy regarding all the
procedures examined.

9. Performance analysis
This section contains the cross validation in the middle of
existing and proposed procedures. Some well-known
image execution parameters for computerized images have
been chosen to demonstrate that the execution of the
proposed algorithm is superior to the current techniques.
9.1 Mean Square Error Evaluation
Mean square error is a measure of image quality record. he
large value of mean square implies that image is a low
quality.
Table.1 Mean Square Error Evaluation
Images
Max-DCT
Proposed
Image1

40

Image2

11

Image3

0.987

Volume 3, Issue 5, September-October 2014

ISSN 2278-6856

Images

Max-DCT

Proposed

Image4

Image5

Image6

11

Image7

0.764

Image8

0.758

Image9

Image10

65

Table 1 is demonstrating the quantized analysis of the

mean square error. As mean square error needs to be
lessened therefore the proposed algorithm is demonstrating
the superior results than the accessible methods as mean
square error is fewer in each case

Fig. 7 MSE of MAX-DCT & Proposed Approach for

different images
Figure 7 has demonstrated the quantized analysis of the
mean square error of distinctive images using fusion by
DWT transform (Black Color), fusion by DCT transform
(Magenta Color), fusion by PCA transform (Blue Color),
fusion by Proposed Approach (Red Color). It is clear from
the plot that there is lessening in MSE value of images
with the use of proposed technique over different
techniques. This diminishing symbolizes improvement in
the goal nature of the image.
9.2 Peak Signal to Noise Ratio Evaluation
The PSNR block calculates the peak signal-to-noise ratio,
between two images. This ratio is frequently used as
quality estimation between the first and a fused image. The
higher the PSNR demonstrates the better the nature of the
fused or reconstructed image.
Table. 2 Peak Signal to Noise Ratio Evaluation
Images
Max-DCT
Proposed
Image1

64.0309

82.4717

Page 5

International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)

Web Site: www.ijettcs.org Email: [email protected]
Volume 3, Issue 5, September-October 2014
Image2

75.2049

77.3510

Image3

93.0782

98.0330

ISSN 2278-6856

Table.3 Normalized Cross-Correlation Evaluations

Images
Max-DCT
Proposed
Image1

0.9569

.99

Image2

0.9986

1.0064

Image3

1.0001

1.0000

Image4

0.9985

1.0003

Image5

0.9994

1.0001

Image4

46.9695

89.6805

Image5

85.3504

90.7525

Image6

54.9511

82.9654

Image7

93.3311

98.5240

Image8

78.1538

98.2394

Image6

1.0174

1.0011

Image9

77.4619

92.7624

Image7

0.9957

1.0001

Image10

39.9786

80.4440

Image8

0.9828

1.0000

Image9

0.9812

1.0030

Image10

1.0704

0.9973

Table 2 is demonstrating the similar analysis of the Peak

Signal to Noise Ratio (PSNR). As PSNR need to be
expanded; so the fundamental objective is to expand the
PSNR however much as could be expected. Table 2 has
clearly demonstrated that the PSNR is most extreme on
account of the proposed algorithm in this manner proposed
algorithm is giving superior outcome than the available
techniques.

Fig. 8 PSNR of MAX-DCT & Proposed Approach for

different images
Figure 8 has demonstrated the quantized analysis of the
peak signal to noise ratio of distinctive images using
fusion by DWT transform (Black Color), fusion by DCT
transform (Magenta Color), fusion by PCA transform
(Blue Color), fusion by Proposed Approach (Red Color). It
is very clear from the plot that there is expand in PSNR
estimation of images with the use of proposed techniques
over different strategies.
9.3 Normalized Cross-Correlation Evaluation
Table 3 is demonstrating the relative analysis of the
Normalized Cross-Correlation (NCC). As NCC needs to be
near to 1, in this way proposed algorithm is demonstrating
superior outcomes than the existing methods as NCC is
near to 1 in every case.

Volume 3, Issue 5, September-October 2014

Figure 9 has demonstrated the quantized analysis of the

Normalized Cross-Correlation of diverse images using
fusion by DWT transform (Black Color), fusion by DCT
transform (Magenta Color), fusion by PCA transform
(Blue Color), fusion by Proposed Approach (Red Color). It
is very clear from the plot that there is estimation of NCC
is near to1 in each case with the use of proposed technique
over different techniques. This represents improvement in
the objective quality of the image.

Fig. 9 NCC of DWT, DCT, and PCA & Proposed

Approach for different images
9.4 Average Difference Evaluation
Table 4 is demonstrating the comparative analysis of the
Average Difference. As Average Difference needs to be
minimized; so the primary idea is to lessen the Average
Difference much as could be expected. Table 6.4 has
clearly demonstrated that Average Difference is less in our
situation hence the proposed algorithm has indicated
critical outcome over the accessible algorithm.

Page 6

International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)

Web Site: www.ijettcs.org Email: [email protected]
Volume 3, Issue 5, September-October 2014

ISSN 2278-6856

Table.4 Average Difference Evaluation

Images
Max-DCT
Proposed

Table.5 Normalized Absolute Error Evaluation

Images
Max-DCT
Proposed

Image1

4.2257

0.8964

Image1

6.3939

2.2118

Image2

2.2715

0.9998

Image2

3.3606

2.9701

Image3

0.3078

.0089

Image3

1.2011

0.1606

Image4

0.0988

0.0589

Image4

1.7073

1.4606

Image5

0.0641

0.0033

Image5

1.8740

1.3732

Image6

7.5781

0.1690

Image6

10.7836

2.1498

Image7

0.5368

0.0090

Image7

1.1838

0.3921

Image8

2.4276

0.0027

Image8

2.8359

0.1998

Image9

1.2457

0.3132

Image9

2.9512

1.2232

Image10

20.0586

0.8356

Image10

25.5314

2.4857

Figure 11 has demonstrated the quantized analysis of the

Normalized Absolute Error of distinctive images using
fusion by DWT transform (Black Color), fusion by DCT
transform (Magenta Color), fusion by PCA transform
(Blue Color), fusion by Proposed Approach (Red Color). It
is clear from the plot that there is diminishing in NAE
estimation of images with the use of proposed system over
different techniques. This reduction speaks to change in
the target nature of the image.

Fig. 10 AD of MAX-DCT & Proposed Approach for

different images
Figure 10 has demonstrated the quantized analysis of the
Average Difference of diverse images using fusion by
DWT transform (Black Color), fusion by DCT transform
(Magenta Color), fusion by PCA transform (Blue Color),
fusion by Proposed Approach (Red Color). It is clear from
the plot that there is reduce in AD estimation of images
with the use of proposed strategy over different system.
This reduction shows enhancement in the objective quality
of the image.
9.5 Root Mean Square Error Evaluation
Table 5 has demonstrated the quantized analysis of the
Root Mean Square Error. As Root Mean Square Error
needs to be decreased in this way the proposed algorithm is
demonstrating the superior outcome than the accessible
techniques as Root Mean Square Error is less in all case.

Volume 3, Issue 5, September-October 2014

Fig. 11 NAE of DWT, DCT, and PCA & Proposed

Approach for different images

10. Conclusion
From the survey it is concluded that the existing
researchers have ignored few of the issues of vision
processing. These issues are related to image de-noising,
image enhancement, and image restoration. So to rise
above these issues, we have proposed a novel algorithm in
this paper. The proposed work incorporate dynamic
histogram equalization with consistency verification based
Max-DCT based fusion technique to give improved results
Page 7

International Journal of Emerging Trends & Technology in Computer Science (IJETTCS)

Web Site: www.ijettcs.org Email: [email protected]
Volume 3, Issue 5, September-October 2014
than the accessible techniques. The proposed technique has
defeated the limitations of the exhibit fusion techniques.
Comparative analysis has demonstrates the enhancement
of the proposed algorithm in terms of the performance
metrics. We can expand this work to use decision based
median filters in near future. This will bring about upgrade
of the proposed max-DCT base fusion in proficient way.
To take the profits of the proposed algorithm, we can
likewise expand this work to use it in smart cameras by
using the embedded systems.

References
[1] Albuquerque, Hugo R., Tsang Ing Ren, and George
DC Cavalcanti. "Image Fusion Combining Frequency
Domain Techniques Based on Focus." Tools with
Artificial Intelligence (ICTAI), 2012 IEEE 24th
International Conference on. Vol. 1. IEEE, 2012.
[2] Drajic, Dejan, and NedeljkoCvejic "Adaptive fusion of
multimodal surveillance image sequences in visual
sensor networks." Consumer Electronics, IEEE
Transactions on 53.4 (2007): 1456-1462.
[3] Vladimir S. Petrovic , "Multisensor Pixel-level image
Fusion," feb
[4] DIP,
[Lat
visited]
27-august-2014,
[ONLINEAVAILABLE]
http://www.edn.com/design/sensors/4397908/Biometri
c-Authentication--Multicue-data-fusion-Part-I
[5] Phamila, Y., and R. Amutha "Discrete Cosine
Transform based fusion of multi-focus images for
visual sensor networks." Signal Processing 95 (2014):
161-170.
[6] Gupta, Vivek Kumar, AmitNeog, and S. K. Katiyar
"Analysis of image fusion techniques over
multispectral
and
microwave
SAR
images." Communications and Signal Processing
(ICCSP), 2013 International Conference on.IEEE,
2013.
[7] Prakash, Om, RichaSrivastava, and AshishKhare
"Biorthogonal wavelet transform based image fusion
using absolute maximum fusion rule." Information &
Communication Technologies (ICT), 2013 IEEE
Conference on.IEEE, 2013.
[8] Hossny, Mohammed, et al "Image fusion metrics:
evolution in a nutshell."ComputerModelling and
Simulation (UKSim), 2013 UKSim 15th International
Conference on.IEEE, 2013.
[9] Desale, Rajenda Pandit, and Sarita V. Verma. Study
and analysis of PCA, DCT & DWT based image
fusion techniques. In Signal Processing Image
Processing & Pattern Recognition (ICSIPR), 2013
International Conference on, pp. 66-69. IEEE, 2013.
[10] Parmar, K., &Kher, R (2012, May). A comparative
analysis of multimodality medical image fusion
methods.In Modelling Symposium (AMS), 2012 Sixth
Asia (pp. 93-97).IEEE
[11] Parmar,kiran,RahulK.Kher, and Falgun N. Thakkar,
Analysis of CT and MRI Image Fusion Using
Wavelet Transform. Communication Systems and

Volume 3, Issue 5, September-October 2014

ISSN 2278-6856

Network Tech nologies (CSNT), 2012 International

Conference on. IEEE.2012
[12] Wang Xin, Li Yingfang A New Method for MultiFocus Image Fusion Using Countourlet Transform
2011 International Conference on Transportation,
Mechanical, and Electrical Engineering (TMEE)
December 16-18, Changchun, China
[13] Patil, Ujwala, and Uma Mudengudi Image fusion
uing hierarchical PCA. Image Information
Processing (ICIIP), 2011 International Conference on.
IEEE,2011.
[14] A. Soma Sekhar, Dr.M.N.Giri Prasad. A Novel
Approach Of Image Fusion On MR And CT Images
Using Wavelet Transforms. IEEE Trans. on Image
Proc., pp. 172-176. IEEE, 2011.
[15] Ahmed Abd-el-kader , Hossam El-Din Moustafa ,
SamehRehan Performance Measures for Image
Fusion Based on Wavelet Transform and Curvelet
Transform 28th NATIONAL RADIO SCIENCE
CONFERENCE(NRSC 2011)April 26-28, 2011,
National Telecommunication Institute, Egypt
[16] Su-xia, Xing, et al Image Fusion Method Based on
NSCT and Robustness Analysis. Computer
Distributed Control and Intelligent Environment
Monitoring
(CDCIEM),
2011
International
Conference on. IEEE, 2011
[17] Ghimire Deepak and Joonwhoan Lee. Nonlinear
Transfer Function-Based Local Approach for Color
Image Enhancement. In Consumer Electronics, 2011
International Conference on, pp. 858-865. IEEE,2011

Page 8