International Journal on Future Revolution in Computer Science & Communication Engineering ISSN: 2454-4248

Volume: 4 Issue: 2 92 – 95
_______________________________________________________________________________________________
Study of Various Image Fusion Techniques for the Enhancement of Digital
Images

Ravneet Kaur Inderdeep Kaur

Research Scholar (A.P CSE & M.Tech Coordinator)
(M.tech Pursuing ,CSE) GGSCMT,Kharar,Punjab,India
GGSCMT,Kharar,Punjab,India Chandigarh, India
ravneetkalyan25@gmail.com kaur.inderdeep@gmail.com

Abstract—Image fusion is process which is applied to combine two distorted images in order to obtain a single meaningful
image. Image fusion is done by extracting some important features from more than one image. Then these extracted features are
merged to a single image which is fused image. The fused image has more enhanced quality in comparison to the input images.
The image fusion is done by using specific techniques. Some of the techniques are discussed in this work. Image fusion attracts
number of researchers to this domain. This work also provides an overview to the work that had been done in past by various
authors. This is a guide to the scholars who are working in this domain.
Index Terms— Image Fusion, Image Enhancements, Transform Domain, Spatial Domain, Frequency Domain.
__________________________________________________*****_________________________________________________

I. INTRODUCTION fusion or enhancement methods are divided in two forms as

below:
With the advancement in the technology each and every
field is modulated to digitalization. Hence the number of 1. Spatial Domain
increased users leads to the induction in problems and errors 2. Frequency Domain
in technology. Hence image fusion is a solution to one of
these problems. The purpose of image fusion is to retain the Spatial domain techniques are meant to deal with the pixels
enrich part of the image. The meaningful information of the in an image. In this the pixel values are altered to perform
image can be maintained by combining the multiple images. the enhancements in the image. The following image fusion
There are no constraints or limitations on domain of the techniques come under this category:
input images. The input image can be related to multi focus, a. Brovey Method
multi sensor or other domains. The image below defines the b. Principal Component Analysis
example of image fusion. As there are two images which are c. HIS based Method
not clearly visible one is blurred from right angle and other
is blurred from left angle. In Frequency domain technique, first of all the image is
transformed to the frequency domain. In this, the Fourier
Image fusion is a technique which generates a more transform is evaluated and then rests of the operations are
enhanced image by combining two or more than two images applied to it. After then the inverse Fourier transform is
into a single. This is done for achieving more enhanced or applied to achieve the output image. The image
meaningful image so that the fused image can lead to enhancement is done to increase the quality of the image by
effective decision making in near future. Generally the making few modifications to its content. The modifications
definition of image fusion is to combine two or more than brightness, contrast, distribution of the grey level etc are
two images in order to produce a single image. The need of possible to perform [14].
image fusion arises when we have more than two images
and all of them lack somewhere hence storing all of such 2.1 Principal Component Analysis
images will consume large memory space and almost all of
the images does not contain the relevant information PCA is conversion method which converts the number of
individually. At last the solution to this problem is associated variables into the format of set of unassociated
developed in the form of image fusion. The process of variable. It works effectively and efficiently in case of
image fusion starts by introducing two images as input and image compression and image classification. The conversion
then by implementing suitable fusion technique these input of associated variables is done with the help of mathematical
images are derived to a single image which contains all of calculations and functions. It leads to the concise and finest
the meaningful content of the image that was original. The description of data set [19]. The function of PCA is
image fusion solves the problem of memory area implemented in the form of principal component. The first
consumption in excess and incomplete meaningless images. principal component examined much of the dissents in the
available data as supple. The second principal component is
II. TECHNIQUES encumbered within the subspace which is perpendicular to
first component. Then the value of maximum variance is
Image fusion techniques are used for enhancing the quality selected as a third component. The PCA is a alternative
of any digital images without ruining its features. The image
92
IJFRCSCE | February 2018, Available @ http://www.ijfrcsce.org
_______________________________________________________________________________________
International Journal on Future Revolution in Computer Science & Communication Engineering ISSN: 2454-4248
Volume: 4 Issue: 2 92 – 95
_______________________________________________________________________________________________
mechanism which can be used in place of HIS technique. The figure below shows that how the image is decomposed
The working of PCA is as follows: in DWT mechanism. First of all, the image is decomposed
into right and left sub-bands. Then these sub-bands are
(1) First step is to electrocute IR to the value of PAN and further divided into upper and lower regions.
MS and also resample the MS.
(2) Conversion of MS bands into components such as PC1,
PC2, and so on.
(3) Twin of histogram between PAN and PC1.
(4) Reconstitute the PC1 with PAN.
(5) Conversion of PAN to left principal components.
The working of the PCA technique is depicted in figure
below.
Figure 2 Discrete Wavelet Transform [10][11]
2.2 Discrete Cosine Transform
The major characteristic of Discrete Cosine Transform is
2.4 Stationery Wavelet Transformation
that it facilitates the good signal approximation on the basis
of coefficient values. This technique is used by many The SWT is a Stationery Wavelet Transformation technique
algorithms for transforming the image and as well as for which is developed to overcome the weaknesses of Discrete
watermarking also. The main advantage of this technique is Wavelet Transformation technique. The problem found in
that it has fast processing speed in contrast to other DWT was lack of translation Invariance the solution of this
techniques. This is the easy and lossless compression is to remove or vanish the down samplers and up samplers
mechanism. in DWT. And perform the up sampling on filter coefficients
in the jth level of algorithm by applying the following
It decomposes the image into two parts such as spectral sub
equation:
parts. The working of this technique is quite similar of DFT
i.e. Discrete Fourier Transformation since it also converts 2 𝑗 −1 (1)
the image to frequency domain from spatial domain.
It is an a redundant technique because the number in the
[3]. each level of output is similar to the number in
DCT
DCT
corresponding input which leads to the N number of
First redundancy corresponding to the N number wavelet
Image Fusion
Rule coefficients. The other name of SWT is “Algorithm of
Fused Trous”. This name is derived from French a term which
Image refers to the insertion of zeros at filters. The meaning of
“Trous” in English is holes. It was developed by
holschneider and also named as undecimated wavelet
transform or algorithm of trous. It is an enhanced or
IDCT

Second modified version of DWT which does not destroy

Image coefficient at every transformation level.
DCT
2.5 Laplacian Pyramid
It works on the basis of pyramid structure. This structure
Figure1 Image Fusion with DCT [10] contains the elements like blocks which are related to the
various levels of the real image. The fragmentation of an
2.3 Discrete Wavelet Transform image into various levels is done by implementing low pass
Discrete wavelet technique is used to transform the discrete filters which recursively filters the low level image. As the
signals into discrete wavelet representation. In this level moves upward the image become minor by preceding
technique, an image is decomposed into coefficients and the length of the image.
then this coefficient compares with the threshold. Value zero III. BACKGROUND
will be assigned to those whose coefficient value is less than
threshold and the coefficient above the threshold value are Phamila and Amutha [1] in their paper shown that multi
encoded with the loss less compression technique. DWT focus image scheme which was based upon the higher
involves decompose, threshold detail coefficients and then highly values alternating coefficients abbreviated as AC
reconstruct. Generalize steps for the implementation of coupled together with Discrete Cosine Transform concluded
DWT is as below: as an unproblematic as well as efficient plan in terms of
pertaining cellular image sensor devices. These devices
Step 1: Create a wavelet lower decomposition by applying were built with the help of learning resource limited and
the DWT mechanism. battery pack driven graphic receptors. Moreover, the
Step 2: Apply fusion methods to each decomposition level. utilization of these devices was in the area of security and
Step 3: Apply Inverse DWT for achieving the original high risk atmosphere such as battlefields. The simulation
image. analysis has performed using the proposed AC-DCT method
93
IJFRCSCE | February 2018, Available @ http://www.ijfrcsce.org
_______________________________________________________________________________________
International Journal on Future Revolution in Computer Science & Communication Engineering ISSN: 2454-4248
Volume: 4 Issue: 2 92 – 95
_______________________________________________________________________________________________
which concluded that it has overcome the limitation of exist when there are too many images but all of them are not
power of low power devices along with efficient actual of good quality and less informative. Thus in such case the
calculation. informative part of each image is combined to a single
image so that the output or fused image can become more
Li and Dong [2] discussed image fusion techniques applied informative or meaningful. Various types of techniques are
on pixel level. In these fusion techniques, the particular used for image fusion. The objective of this is to enhance
processing along with synergistic combination of the performance of image fusion technique by introducig the
information was performed. The information i.e. combined selective Minimum/Maximum and PCA specifically for
was collected from the source images which focused on spatial domain based images. the results show that the
improving the perception of a scene. With the development proposed spatial domain based technique is better than the
of sensors, demand for correct combination of all available traditional spatial domain based transformation techniques.
image datasets in a significant and spatial manner has been
arisen. The image fusion techniques based on pixel level is Rohan Ashok Mandhare(2013), [7] remote sensing is a
applicable to several areas such as space borne remote technology which is used to capture the image of earth by
sensing, machine vision and airborne. using satellites. Image capturing in remote sensing is done
by using multiple sensors. The images taken by the remote
The work performed by Sharmila et al. in the year of 2013 sensing technology is prone to impulse noise. Hence there
has shown that multimodality medial image fusion using the are many techniques which can be used for removing the
wavelet transformation and entropy produced better quality noise from the images and named as image filtering
results in terms of quality information and less noise in the techniques.The objective of this research work was to
produced image. In the medial image fusion, the primary implement pixel level image fusion based on mathematical
data can be derived simply using incorporating and wavelet based transformation method for image fusion.
multimodality medical images such as Computed The improvement in the capacity of spectral and spatial
Tomography, Positron Emission Tomography, Magnetic information was also evaluated. Various methods were used
Resonance Imaging and single Photon Emission Computed for practical implementation and the objectives was
Tomography into a single image. This derived information achieved by using various methods such as average method,
can be used for several reasons including detection of tumor, wavelet based transformatuion, multiplicative method and
diagnoses of diseases, surgery treatment etc. On the brovey method. The performance evaluation was done by
counterpart, Single modality image is not capable enough in using various performance parameters suh as PSNR,
providing useful information. entropy, standard deviation and RMSE. The results depicted
Anju Rani implemented image enhancement using image that the multiplicative method and wavelet based method
fusion technique in [4]. Image enhancement is a process achieves high efficiency in compaison to other methods. The
which makes an image more meaningful and informative. In fused image generated by these methods also has higher
this work image fusion was used for enhancing the image spectral efficiency.
quality. As the result of image enhancement the generated In the year of 2013, Liu et al. [8] unveiled that lifting
output image can be used for various purposes in near future schemes can be implemented to the multi focus image
by human being as well as machine. This paper provided a fusion for achieving the substantial information in final
summarized review of various image enhancement image. This technique processed in a high speed and
techniques which were combined with image fusion consumes less storage. Thus these features make it more
techniques. A comparison between different image preferable over rest of the techniques. This method divided
processing techniques was also explained by author. the input images into four sections which is known as sub
Vadher Jagruti [5] in 2014 defined that with the bands i.e. LL, HL, HH and diagonal. These bands were
advancement in the technology the working process of each comprised of frequency information of the image. Gaussian
and every area has been enhanced to automatic system from Kernel was implemented for ascertaining the weighted area
mannual systems. Hence various fields use the multisensor energy. After observing the highest energy level between
data. Due to which the need of image fusion arises. The two images, it generated a binary map. Then sub band
image fusion is used for combining two or more images to synthesizing was performed to obtain the final output image.
generate a single meaningful image. This study provided a IV. CONCLUSION AND FUTURE SCOPE
new hybrid technique for multi sensor image fusion by
combining two prominent techniques i.e. DCT and Image fusion is implemented with an objective to integrate
Variance. Traditionally DWT was used for image fusion two distorted image with a purpose to achieve a single but
because it is simple and less complex to execute. The DWT meaningful image. The fusion is done to enhance the
is a robust and a flxible technique since it supports simple quality of an image. It is done by implementing different
transformation amid cupious multi scale transform. It is fusion techniques such as PCA, SWT, DWT, DCT etc. After
quite analogous to wavelet based image fusion techniques. having a review to the previous work it is concluded that it
The comparison was done on the basis of hybrid DWT is required to optimize the fused results by applying various
method for image fusion. optimization techniques.
In 2014, C. Morris [6] explained that the image fusion is a
process in which a set of images is collaborated as a single
image. The purpose of image fusion is to make an image
more enhanced and informative. The need of image fusion
94
IJFRCSCE | February 2018, Available @ http://www.ijfrcsce.org
_______________________________________________________________________________________
International Journal on Future Revolution in Computer Science & Communication Engineering ISSN: 2454-4248
Volume: 4 Issue: 2 92 – 95
_______________________________________________________________________________________________
V. REFERENCES International Journal of Computer Science (IIJCS) Volume 2,
Issue 4, April 2014.
[1] Y. Asnath Victy Phamila and R. Amutha. “Discrete Cosine [12] Huaxun Zhang and Xu Cao. “A way of image fusion based on
Transform based fusion of multi-focus images for visual wavelet transform” In 9th International Conference on Mobile
sensor networks.” Signal Processing 95 (2014). Ad-hoc and Sensor Networks, IEEE, 2013.
[2] Mingjing Li and Yubing Dong. “Review on technology of [13] Gazal Malhotra and Dr. Vinay Chopra. “Improved multi-focus
pixel-level image fusion.” In International Conference on image fusion using ac-dct, edge preserving smoothing &
Measurement, Information and Control (ICMIC),vol. 1, IEEE, DRSHE” In Proceedings of International Conference on
2013. Computer Science, Cloud Computing and Applications July
[3] K Sharmila, S Rajkumar, V Vijayarajan. “Hybrid method for 24-25, 2014.
Multimodality Medical image fusion using Discrete Wavelet [14] Sweta k. Shah and Prof. D.U. Shah, “A Review on Image
Transform and Entropy concepts with Quantitative Analysis.” Fusion Techniques”, International Journal Of Innovative
In International conference on Communication and Signal Research In Electrical, Electronics, Instrumentation And
Processing (ICCSP), IEEE, April 3-5, 2013. Control Engineering, Vol. 2, No. 3, Pp. 1258-1261, March
[4] Anju Rani et al (Sep 2014), “Image Enhancement using Image 2014
Fusion Techniques”, IJARCSSE, Vol. 4, No. 9, Pp. 413-416 [15] Shalima and Dr. Rajinder Virk, “Review Of Image Fusion
[5] Vadher Jagruti (2014),”Implementation of Discrete Wavelet Techniques”, International Research Journal of Engineering
Transform Based Image” Fusion. Journal of Electronics and and Technology, Vol. 2, No. 3, Pp. 333-339, June 2015
Communication Engineering ,Volume 9, Issue 2, PP 107-109 [16] Er.Kulvir Singh and Er.Neeraj Julka, “A Review on Image
[6] C. Morris(2014), “two stage spatial domain image fusion Fusion Techniques and Proposal of New Hybrid Technique”,
techniques”, ICTACT journal on image and video processing, International Research Journal of Engineering and
volume: 05, issue: 01 , pp8895-898, Technology, Vol. 3, No. 3, Pp. 1321-1324, March 2016
[7] Rohan Ashok Mandhare(2013), “2690 Pixel-Level Image [17] Deepak Kumar Sahu and M.P.Parsai, “Different Image Fusion
Fusion using Brovey Transform and Wavelet Techniques –A Critical Review”, International Journal of
Transform”,International Journal of Advanced Research in Modern Engineering Research, Vol. 2, No. 5, Pp. 4298-4301,
Electrical, Electronics and Instrumentation Engineering October 2012
(IJAREEIE ), Vol. 2, Issue 6, pp2690-2695 [18] Suruchi Goyal and Rupinder Wahla, “A Review on Image
[8] Qingping Li, Junping Du, Fuzhao Song, Chao Wang, Fusion”, International Journal of Innovative Research in
Honggang Liu, Cheng Lu. “Region-based Multifocus Image Computer and Communication Engineering, Vol. 3, No. 8, Pp.
Fusion Using the Local Spatial Frequency”, In Control and 7582-7588, August 2015
Decision Conference (CCDC), 25th Chinese, IEEE, 2013. [19] Dhirendra Mishra and Bhakti Palkar, “Image Fusion
[9] L Ganesh, S P Krishna Chaitanya, J Durga Rao and M N V S Techniques: A Review”, International Journal of Computer
S Kumar. “Development of Image Fusion Algorithm for Applications, Vol. 130, No. 9, Pp. 7-13, November 2015
Impulse Noise Removal in Digital Images using the quality
Assessment in Spatial Domain” In International Journal of
Engineering Research and Applications (IJERA) Vol. 1, Issue
3, 2007.
[10] Lixin Liu, Hongyu Bian and Guofeng Shao. “An Effective
Wavelet-based Scheme for Multi-focus Image Fusion” In
IEEE International Conference on Mechatronics and
Automation (ICMA), 2013.
[11] Jasmeet kaur and Er. Rajdavinder Singh Boparai. “An
evaluation on different image fusion techniques” IPASJ

95
IJFRCSCE | February 2018, Available @ http://www.ijfrcsce.org
_______________________________________________________________________________________