Fichier PDF

Partage, hébergement, conversion et archivage facile de documents au format PDF

Partager un fichier Mes fichiers Convertir un fichier Boite à outils PDF Recherche PDF Aide Contact



2015021112wavelets .pdf



Nom original: 2015021112wavelets.pdf
Auteur: paresh

Ce document au format PDF 1.5 a été généré par Microsoft® Office Word 2007, et a été envoyé sur fichier-pdf.fr le 14/11/2015 à 19:05, depuis l'adresse IP 41.150.x.x. La présente page de téléchargement du fichier a été vue 243 fois.
Taille du document: 350 Ko (4 pages).
Confidentialité: fichier public




Télécharger le fichier (PDF)









Aperçu du document


International Journal For Technological Research In Engineering
Volume 2, Issue 11, July-2015

ISSN (Online): 2347 - 4718

MEDICAL IMAGE FUSION USING WAVELET TRANSFORM
Manpreet Kaur1, Dr. Shiv Kumar Verma2, Gagandeep Kaur3
1,3
PG Research Scholar, 2Associate Professor
Department of CSE, Chandigarh University, Gharuan, Punjab
Abstract: Medical image fusion contain the complementary
and relevant information from multiple source images that
used for identify the diseases and better treatment. Fusion
can be done either at pixel level or feature level and or
symbol level. Spatial and Transform are two basic
approaches for fusion of image. The available technique for
fusion of multiple images into single image are; Virtual
Navigator Technology, Volume Delineation Technique,
Novel MIF method, Discrete Transform Wavelet and
Undecimated Wavelet Transform. Advance wavelet families
are Contourlet transform, Curelet transform and Non
Subsample Contourlet transform that provide better
performance, these transform are high cost and consume
lot of memory space. In this paper complete functional
survey of all techniques for fusion has been described and
work done using Discrete Wavelet Transform due to low
cost and less overhead also important feature of this
transform are locality, multi resolution analysis, edge
detection , decor relation and energy compaction.
Keywords: Medical image Fusion, levels, Methods, Wavelet
Transform, Discrete Wavelet Transform.
I. INTRODUCTION
Medical image is important part in human health care field
that provide information about human body structure for
better treatments. CT scan is provide bones structure of
human body, MRI image give soft tissues information and
PET image gives information about flow of blood in human
body with low spatial resolution. Fusion is a process in which
two or more images are combined to get new one image as a
whole is called fused image. Fused image contains more
relevant, accurate information than input images. The
medical image fusion is important part of medical field to
integrate two or more human body images into one to
identify all possible diagnosis, diseases in human body so
that effective treatment applied. The basic requirements for
image fusion are that input images contain relevant
information that are required in fused image and fusion
process does not introduces any distortion, unwanted features
in fused image [8]. Medical image fusion is defined into three
types: Pixel level fusion, Feature level fusion and Decision
level fusion. Pixel level fusion [2] is concerned with
information about each pixel of input image and fusion image
is obtained with respect to pixel values of input images. In
feature level fusion [5] is divide input image into regions or
features (pixel intensities, textures, edges) to get fusion
image. Decision level fusion is used voting, fuzzy logic [11]
for fusion. Spatial and transform are two basic domains [10]
for fusion of input images. In spatial domain fusion is
directly worked on intensity (pixel) of input image. Principal

www.ijtre.com

component analysis (PCA) [13] [14], linear fusion, and sharp
fusion are some examples of spatial domain of fusion. Sharp
spatial domain contain spatial distortion problem that was
noted in paper [13].Due this spatial domains schemes are not
normally used in medical image field. Therefore transform
domain is commonly preferred in medical image field to
remove problem of spatial domain. Transform domain
contain pyramid and wavelet transforms. Some pyramid
transforms are Laplace pyramid [1], gradient pyramid [6]
and contrast pyramid. Wavelet transform is used rather than
Pyramid transform due to blocking problem. Discrete
wavelet transform [10] is powerful method used in medical
image fusion. This type of transform provides better, relevant
information with low cost and less overhead. The other
important advantages of discrete wavelet transform are
locality, edge detection, energy compaction, decor relation
and multi resolution analysis. In [1], the performance of
wavelet transform, HIS, PCA fusion techniques were
compared. The fused image is obtained with different RGB
values for different fusion techniques. In wavelet transform,
different images are accurately combined by using inverse
wavelet transform of fused wavelet coefficients. HIS
Transform is used for combining multi-resolution. In [3]
different fusion levels have been defined but only pixel level
fusion is used for fusion purposes. The resulted image is
improved by using undecimated wavelet transform that
divide image into filtering operations by using analysis filter.
The orthogonal filters are developed for image fusion and
provide useful properties such as a short support size of the
analysis filter. By using orthogonal filter, the unwanted
spreading of coefficient values around image is decreased
and unwanted objects in the fused reconstruction is reduced,
while [4] discussed HIS technique for sharpening of images,
which is used for enhancement of features, color and good
spatial resolution of fusion image. There are different HIS
transformation algorithms used for transferring color image
from RGB form to HIS form. To get fusion of high
resolution, it is important to do treatment in HIS space. In [7]
different image fusion techniques (simple average, simple
minimum, simple maximum, PCA, DWT) were described. A
new approach is present after the comparison of these fusion
techniques. The Wavelet transforms provide a high quality
spectral data means better quality of fused image. DWT and
PCA are combined that provide much better performance
and improve quality of the fused image.In [12] different
fusion techniques like wavelet transform, HIS transform,
Principal component (PC) method has been discussed. Also
different fusion techniques are applied to input image and
their performance were analyzed by using evolution
procedures e.g. SSIM, CC, PD, RMSE, HCC, ED. But

Copyright 2015.All rights reserved.

2525

International Journal For Technological Research In Engineering
Volume 2, Issue 11, July-2015
SSIM, CC, PD, RMSE, shows the defect when multi-sensor
and multi-temporal fusion is used. The methods that identify
fusion quality were tested for multiple images and test sites.
Some tested methods perform well but inconsistent occurred
with visual analysis results. Principal component analysis
(PCA) [14] is a mathematical tool which converts a number
of correlated variables into a number of uncorrelated
variables. The PCA is mostly preferred in image compression
and image classification. In[7] pixel level fusion was
described that combined low spectral-high spatial resolution
content with high spectral-low spatial resolution content The
HPF method used three parameters: high pass filter size,
center value of filter kernel, injection weight of image that
is converted into high pass filters size form. The
improvement version of HPFA method was represented. The
purposed method of this paper is focused on converting
original multispectral properties of input images into high
degree.
II. METHODOLOGY
There are various techniques used in medical image fusion.
On basis of reviewing various papers, we are preferred
discrete wavelet transform for fusion purpose. There are
various advanced wavelet transform e.g. contourlet transform
[8], curvelet transform, and non subsample contourlet
transform which are also used in medical image fusion, gives
better performance of fused image. These transforms are very
expensive and dot not commonly used for fusion purpose.
Discrete transform is easy and efficient method for fusion. It
provides the relevant information of input image with low
cost and less complexity. We used DWT transform [10] for
multiple medical image fusion and developed new multi
scale fusion scheme with multiple selection rules. This new
technique provides the flexibility to choose accurate fused
image. The higher the scale of input image, more detailed
information is obtained in fused image.
To apply DWT Algorithm:
 Input images (CT scan image, MRI image) are
selected that are to be fused in one image.
 Input images contains RGB layers that to be
separated to get intensity of each color in input
images.
 Then apply DWT to each layer of input images.
 Apply inverse DWT method to get fused image
contain best features of both input images.
The DWT [10] of a present signal 𝐹(𝑥) is performed by
using scaling function 𝜙(𝑥) and wavelet function 𝜓(𝑥) to
analysis and synthesis of given signal. The general equation
of multi resolution theory is scaling equation i.e. given as:
∅ 𝑥 = √2 𝑙 𝑘 𝜙 2(𝑥 − 𝑘) ………. (1)
Where 𝑙 𝑘 is low-pass coefficients and √2 controls norm of
scaling factor.
LL2
LH2
LH1
HL2
HH2
HL1
HH1
Fig.1 Discrete wavelet transforms (DWT) decomposition

www.ijtre.com

ISSN (Online): 2347 - 4718

process up to level 2.
The wavelet function 𝜓(𝑥) that computes detailed
coefficient which is given by
𝜓 𝑥 = √2 𝑘 ℎ 𝑘 𝜙(2𝑥 − 𝑘) ………. (2)
Where ℎ(𝑘) is high frequency or detailed wavelet
coefficient.
Scaling coefficients 𝑙(𝑘) and wavelet coefficients ℎ(𝑘) are
performed for signal divisions.

Fig. 2 General representation of fusion process in wavelet
Transform
After this step, reconstruction of signals is done by
combining scaling coefficients and wavelet coefficients that
is represented as
𝐶 𝑗 + 1, 𝑘 = 𝑘 𝐶 𝑗, 𝑘 𝑙 𝑚 − 2𝑘 + 𝑘 𝐷 𝑗, 𝑘 ℎ(𝑚 −
2𝑘)……………. (3)
Where 𝐶 𝑗, 𝑘 are scaling coefficients and 𝐷 𝑗, 𝑘 are
wavelet coefficients.
The fusion technique that we are followed is based on
maximum selection scheme means maximum wavelet
coefficients values that carry other features like edge,
boundaries, and contours. Therefore value of wavelet
coefficients has been used selecting fused wavelet
coefficients. Let there are two source medical images
𝑠1(𝑥, 𝑦) and 𝑠2 𝑥, 𝑦 , the procedure of fusion approach is
written as:
Separation of sources images using DWT:
𝑊1𝑙 (𝑥, 𝑦) = 𝐷𝑊𝑇[𝑠1(𝑥, 𝑦)]
𝑊2𝑙 (𝑥, 𝑦) = 𝐷𝑊𝑇[𝑠2(𝑥, 𝑦)]
𝑙
Where 𝑊1 (𝑥, 𝑦) and are wavelet coefficients of source
image 𝑠1(𝑥, 𝑦) and source image 𝑠2 𝑥, 𝑦 at scale l.
Measuring fused wavelet coefficients at scale l:
𝑊1𝑙 𝑥, 𝑦 , 𝑖𝑓 𝑊1𝑙 𝑥, 𝑦 > 𝑊2𝑙 𝑥, 𝑦
𝑊𝐹𝑙 𝑥, 𝑦 =
𝑊2𝑙 𝑥, 𝑦 , 𝑖𝑓 𝑊2𝑙 𝑥, 𝑦 > 𝑊1𝑙 𝑥, 𝑦
Reconstruct fused image 𝐹 𝑙 𝑥, 𝑦 at scale 𝑙 using inverse
DWT:
𝐹 𝑙 𝑥, 𝑦 = 𝐼𝐷𝑊𝑇[𝑊𝐹𝑙 (𝑥, 𝑦)]

Copyright 2015.All rights reserved.

2526

International Journal For Technological Research In Engineering
Volume 2, Issue 11, July-2015

ISSN (Online): 2347 - 4718

This method performs efficiently and effectively may
identify the anomalies exists in an objects to detect and
analyzed for radiological purpose. Still more work are
needed to be done for enhancement in the technique make it
better compare to other techniques.

Fig.3 Fusion Process
This fused technique is compared with other fusion methods
i.e. GP, PCA, RP and CP that do not provide the information
from CT scan MRI pairs. This proposed approach gives
better quality than DWT with DBSS.DWT transform method
is easily implemented as compared to other method of fusion.
According to the requirement, the levels of decomposition
process can be increased.
III. RESULTS AND COMPARISON
Every fusion method gives different performance and
different results.DWT transform method is based upon the
frequency level fusion, where as spatial Transform is based
on pixel level fusion. In [9], DWT was tested with different
fusion rules i.e. maximum rule, minimum rule, mean
rule.DWT with maximum rule was selected due its better
results. The PCA technique has spectral degradation problem,
Contourlet Transform is still based on pixel level fusion and
very expansive. We used DWT transform due its easy
implementation, less expansive as compared to other
advanced wavelet transforms. The following table [10]
showed different results with different techniques such as
Gradient pyramid (GP), Contrast pyramid (CP), Ratio
pyramid (RP), Principal Component Analysis (PCA).
Table1: Quantitative evaluation of fusion results:
Fusion
Q
MI
E
SD
SF
method
GP
0.5784 1.0243
5.4698 19.7350 6.6479
CP
0.2542 0.9452
1.9243 33.3962 14.5826
RP
0.2658 0.9901
3.5655 33.1739 14.6512
PCA
0.6395 2.6305
5.6220 28.3806 6.9945
PM
0.7286 2.1396
5.9335 32.9163 9.9138
Q - Edge Strength, MI - Mutual Information, E - Entropy’s,
SD - Standard Deviations, SF - Spatial Frequency, and PM –
Proposed Method.
IV. CONCLUSION
DWT method provides more relevant detailed information in
fused image. It has spatial resolution problem during fusion
process. The proposed technique for the multi-resolution
fusion approach using DWT method exhibits better results.

www.ijtre.com

REFERENCES
[1] P. J. Burt and E. H. Adelson, “The laplacian
pyramid as a compact image code,” IEEE
Transactions on Communications, V (31), pp. 532–
540, 1983.
[2] A. Ellmauthaler , E. A. B. daSilva, C. L. Pagliari
and M. M. Perez, “Multiscale Image Fusion Using
the Undecimated Wavelet Transform With NonOrthogonal Filter Banks”,2012.
[3] Abhishek
Singh,M.Saini
,P.Nayyer,
“Implementation & comparative study of different
fusion techniques (WAVELET, IHS, PCA)”,
International Refereed Journal of Engineering and
Science (IRJES) Volume (4), pp.37-41, 2012
[4] F. A. Al-Wassai, N.V. Kalyankar, A. A. Al-Zuky,
“The IHS Transformations Based Image Fusion”,
Computer Vision and Pattern Recognition (cs.CV)
,19 July 2011
[5] J. J. Lewis ,R. J. O’Callaghan S. G., Nikolov, D. R.
Bull, C. N. Canagarajah, “Region-Based Image
Fusion Using Complex Wavelets”, The Centre for
Communications Research University of Bristol
Bristol , BS8 1UB UK.
[6] J. Tian , Li Chen , L. Ma , W. Yu , “Multi-focus
image fusion using a bilateral gradient-based
sharpness criterion”, Elsevie,2010.
[7] K. Rani , R.Sharma, “Study of Different Image
fusion Algorithm”, International Journal of
Emerging Technology and Advanced Engineering,
V(3), 5 May 2013.
[8] L. Yang, B.L. Guo, W. Ni, “Multimodality medical
image fusion based on multi scale geometric
analysis of contourlet transform”, Published by
Elsevier,2008.
[9] P. Phanindra, J. C. Babu, V. U. Shree, “FPGA
Implementation of Medical Image Fusion Based on
DWT”, International Journal of Advanced Research
in Computer Science and Software Engineering, V
(3), 2013.
[10] R.Singh and A. Khare, “Multi scale Medical Image
Fusion in Wavelet Domain”, Hindawi Publishing
Corporation the Scientific World Journal, 10 pages,
Volume 2013.
[11] S. Das and M. K. Kundu, “A Neuro-Fuzzy
Approach for Medical Image Fusion”, IEEE
Transaction on Biomedical Engineering (60), pp-12,
2013.
[12] S.Klonus, M.Ehlers,“Performance of evaluation
methods in image fusion”,
International
Conference on Information Fusion Seattle, WA,
USA, 2009.
[13] U.G.Gangkofner,
P.S.Pradhan,
and
D.W.

Copyright 2015.All rights reserved.

2527

International Journal For Technological Research In Engineering
Volume 2, Issue 11, July-2015

ISSN (Online): 2347 - 4718

Holcomb,“OptimizingtheHigh Pass Filter Addition
Technique for Image Fusion”, Photogrammetric
Engineering & Remote Sensing V(74),No.9, pp.
1107–1118,2008.
[14] V.P.S. Naidu and J.R. Raol,“Pixel-level Image
Fusion using Wavelets and Principal Component
Analysis”, Defense Science Journal, V (58), No. 3,
pp. 338-352, 2008.

www.ijtre.com

Copyright 2015.All rights reserved.

2528


2015021112wavelets.pdf - page 1/4
2015021112wavelets.pdf - page 2/4
2015021112wavelets.pdf - page 3/4
2015021112wavelets.pdf - page 4/4

Documents similaires


Fichier PDF 2015021112wavelets
Fichier PDF a robust region based multiscale image fusion scheme for mis
Fichier PDF 03 fourier ma
Fichier PDF projet de 3e annee giraud remi 1
Fichier PDF foreground background segmentation using temporal and spatial
Fichier PDF lidar topographie


Sur le même sujet..