Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Abstract-This paper presents fusion of biomedical images based on UDWT and PCA which
aims to integrate the complementary information from multimodal images providing more
information than any of the individual input images. The combination of UDWT and PCA
reduces the redundancy and removes artifacts. The images are fused by applying a specific
fusion rule to improve the image contents. Experimental result shows that our fusion
method gives better performance than existing methods.
Key Words: UDWT, PCA, Entropy, PSNR, Standard Deviation.
I. INTRODUCTION
The common imaging modalities used in medical field are Computed Tomography (CT),
Magnetic Resonance Imaging (MRI), Positron Emission Tomography (PET) etc. CT gives details
about bone structures and does not give information about the soft tissues. But MRI gives
information about soft tissues. The fusion of CT and MRI images provide information about both
bone structures and tissues [1]-[3]. The two approaches to image fusion are Spatial Fusion (SF)
and Transform Fusion (TF).
Discrete Wavelet Transform (DWT) is one of the image fusion methods due to its
simplicity [4]-[7]. But it has some drawbacks namely Shift variance, limited directionality and
absence of phase information. Shift variance can be overcome by the introduction of
Undecimated Wavelet Transform (UDWT). One of the spatial fusion methods is Principal
Component Analysis (PCA) [8]-[11]. Image fusion based on PCA improves the resolution of
images and reduces the redundancy of the image [12]-[[14].Thus UDWT and PCA are combined
in order to increase the contrast and morphological details of an image.
II. IMAGE FUSION BASED ON UDWT AND PCA
The input source images are first decomposed using UDWT. The resultant coefficients
are applied with Principal Component Analysis (PCA) followed by a fusion rule. PCA is an
orthogonal transform and it reduces the redundancy in both the input images. In the next step,
the final image is reconstructed using inverse UDWT.
Input
Image(MRI))
UDWT
PCA
TT
Inverse
Fusion Rule
Input
Image(CT)
UDWT
UDWT
Fused
Image
PCA
(1)
(2)
(3)
(4)
where, (x) is the wavelet function and (x) is the scaling function.
2. Principal Component Analysis (PCA):
PCA is a unique technique in which number of correlated variables is transformed into
number of uncorrelated variables called principal components. The weights for each source
image are obtained from the eigenvector corresponding to the largest Eigen value of the
covariance matrices of each source [16]-[17]. The PCA based image fusion consists of the
following steps:
(a) Arrange the two input images in two column vector. Let S be the resulting column
(b)
(c)
(d)
(e)
vector of dimension 2 n .
Compute the empirical mean M along each column.
Obtain the difference between the mean and S producing the resultant matrix X.
Find the covariance matrix C of matrix X.
C= XXT
(5)
Compute the eigen vectors V and eigen values D of the covariance matrix C and
(6)
(7)
(8)
technique as discussed in this section. The simulation results show that the fused image gives
more information than any one of the input source images. Our present work (i.e.) the
combination of UDWT and PCA for image fusion is compared with other existing image fusion
methods. The performance is studied using different performance metrics such as Entropy (E),
Standard Deviation (SD) and Peak Signal to Noise Ratio (PSNR).
(a)MRI image
(b) CT
image
Fig.2. Input images of different modalities
Input images from different modalities are shown in figure 2. Fig 2(a) represents input
MRI image.Fig 2(b) represents CT image in which the brighter or the white portion of the image
represents the presence of denser matter. First the input images are decomposed using UDWT.
Then the approximation as well as detail coefficients of both decomposed images are fused using
PCA. Next the images are reconstructed using inverse UDWT. Then the performance metrics are
calculated using Entropy(E),Standard Deviation(S.D) and Peak Signal to Noise Ratio(PSNR).
(a)
DWT
(b) PCA
(c) UDWT
d) DWT+PCA (e) UDWT+PCA
Fig. 3 Comparison of fused images
Fig.3 represents that the visual information in the pixels of the fused image using
the proposed method is more than that of the other fusion methods. The quality of the
fused image is analyzed by utilizing suitable objective quality metrics such as Entropy,
PSNR and Standard Deviation. Table I shows the comparison results of the different
fusion methods. It is found that the value of Entropy, Peak Signal to Noise Ratio (PSNR)
and Standard Deviation is high for the proposed method. This signifies that the resultant
fused image obtained using the proposed technique provides more average information.
IV.PERFORMANCE METRICS
L1
P
i=0
log Pi
(10)
f ( n ,m ) ) 2 1/2
(c) Standard Deviation (SD) = { mn (
}
(12)
Higher value of standard deviation indicates better image quality.
TABLE I COMPARISON OF VARIOUS FUSION METHODS USING
FUSION METRICS
Metrics
Fusion Methods
DWT
PCA
UDWT
DWT+PCA
UDWT+PCA
Entropy
6.5772
6.4789
6.6702
6.7504
7.0
PSNR
11.1502
18.4629
12.6155
12.6347
19.2504
Standard Deviation
46.8209
46.2498
48.3422
91.6613
95.7227
Alex Pappachen and Belur V. Dasarathy, Medical image fusion: A survey of the
and
DenoisingInternational
Journal
of
Engineering
Science
Invention,Vol.2,pp.18-27.
[3] Mirajkar Pradnya P,Sachin D. Ruikar,Image Fusion based on Stationary Wavelet
Transform,International Journal of Advanced Engineering Research and Studies,E-ISSN
2249-8974.
[4] Harpreet Kaur,Rachna Rajput,A Combined Approach using DWT & PCA on Image
Fusion,International Journal of Advanced Research in Computer and Communication
Engineering,Vol 4,Issue 9,2015.
[5] Sejal Baraiya, Professor Lokesh P.Gagnani,An Introduction of Image Fusion
Techniques,International Journal for Innovative Research in Science & Technology,
Vol 1,Issue 7,2014.
[6] Bhavana.V,Krishnappa.H.K,Multi-modality medical image fusion using Discrete
Wavelet Transform,Procedia Computer Science,pp.625-631,2015.
[7] Anoop Suraj.A,Mathew Francis,kavya T.S,Nirmal T.M,Discrete Wavelet Transform
based image fusion and denoising in FPGA,Journal of Electrical Systems and
Information Technology,pp.72-81,2014.
[8] Changtao He,Quanxi Liu,Hongliang Li,Haixu Wang,Multimodal medical image
fusion based on HIS and PCA,Procedia Engineering 7,pp.280-285,2010.
component
analysis,Elsevier,Pattern
Recognition
Letters
34,pp.1001-
1008,2013.
[10] Zhijun Wang,Djemel Ziou,Costas Armenakis,Deren Li,Qingquan Li,A Comparative
analysis of image fusion methods,IEEE transactions on Geoscience and Remote
sensing,Vol.43,no.6,june 2005.
[11] Andreas Ellmauthaler,Carla L.Pagliari,Eduardo A.B da Silva,Multiscale image
fusion using the Undecimated
for
medical
images,International
Journal
of
Electronics
and
Communications,pp.896-902,2015.
[15] Prof.Meenakshi M.Pawar,Neeta P.Kulkarni,Image resolution using multi wavelet
transforms with interpolation technique,Journal of Electrical and Electronics
Engineering,Vol.9,pp.09-13,June 2014.
[16] Dr.S.S.Bedi,Rati Khandelwal,Contrast enhancement for PCA fusion of medical
images,Journal of Global Research in Computer Science,Vol.4,March 2013.
[17] Souparnika Jadhav,Image fusion based on wavelet transform,International Journal
of Engineering Research,Vol.3,pp.442-445,July 2014.