Professional Documents
Culture Documents
HTTPS://SITES.GOOGLE.COM/SITE/JOURNALOFCOMPUTING/
WWW.JOURNALOFCOMPUTING.ORG 170
Index Item – Image Fusion, Principal Component Analysis, Wavelet Transform, Luminance, Contrast, Correlation
Coefficient, Entropy, Mutual Information.
Image fusion techniques fall into two groups ‐ i.
2. PRINCIPAL COMPONENT ANALYSIS
Discrete wavelet transform based and ii. Statistical
PCA is mathematically defined as an orthogonal
based. In Statistical based image fusion techniques
linear transformation that transforms the data to a
there are various techniques such as principal
new coordinate system such that the greatest variance
component analysis (PCA) based and histogram
by any projection of the data comes to lie on the first
(HIS) transform based. There are also other image
coordinate (called the first principal component), the
fusion methods like Laplacian Pyramid Method.
second greatest variance on the second coordinate,
This paper is organized in the following way: In and so on. PCA is theoretically the optimum
Section 2 we discuss the fundamentals of the PCA transform for a given data in least square terms.
algorithm. Section 3 briefs our proposed algorithm
For a data matrix, XT, with zero empirical mean
which is the modified version of the traditional PCA
(the empirical mean of the distribution has been
———————————————— subtracted from the data set), where each row
Amit Kumar Sen Assistant Professor, Information Technology
Department, IMPS College of Engineering & Technology. represents a different repetition of the experiment,
Subhadip Mukherjee. and each column gives the results from a particular
Amlan Chakrabarti Assistant Professor, A.K.Choudhury, School of
Information Technology, University of Calcutta. probe, the PCA transformation is given by:
JOURNAL OF COMPUTING, VOLUME 3, ISSUE 4, APRIL 2011, ISSN 2151‐9617
HTTPS://SITES.GOOGLE.COM/SITE/JOURNALOFCOMPUTING/
WWW.JOURNALOFCOMPUTING.ORG 171
Step 2: Generation of data vectors for window
blocks:
The row and the column of every window block
Where the matrix Σ is an m‐by‐n diagonal matrix is arranged to create data vector, i.e. for n window
with nonnegative real numbers on the diagonal and block for first image creates the data vector X1, X2,…,
W Σ VT is the singular value decomposition (svd) of Xn and for second image creates the data vector Y1,
X. Y2,…, Yn.
3. MODIFIED PCA BASED IMAGE FUSION Step 3: Finding the covariance matrix for the
ALGORITHM window blocks.
PCA is a way of identifying patterns in data, and
The covariance between the two images is
expressing the data in such a way as to highlight their
calculated by means of covariance matrix matrix (C)
similarities and dissimilarities. PCA fusion rule is, to
from the image data vectors of Step2. For the ith
find the principal axis Eigen value of the
window block of both the images the covariance is
approximation images, calculate the corresponding
calculated as follows:
eigenvector, and the perform fusion on these
approximation images according to the principal
C =
eigenvector. But there is a disadvantage in these
traditional PCA based image fusion. In traditional
Step 4: Determining the eigenvectors and the
PCA based algorithm [1] it may happen that all the
principal eigenvector.
principal components are selected from the same
region of the image.
The Eigen value of all the Eigen vectors are
calculated from the covariance matrix. The
This drawback is taken care in our proposed
eigenvector that has the maximum value for each of
modification of the PCA algorithm. Our technique is
the window blocks is called the principal Eigen
a window based approach over the existing PCA.
vector. i.e. the principal Eigen vector for all window
Here first we divide the images into some static
blocks are (x1,y1)T, (x2,y2)T … (xn,yn)T.
window blocks. Then we find the principal
eigenvector for each window block and perform Step 5: Calculation of the approximate weight
fusion on two corresponding window blocks of the for every window block.
two images to be fused. This assures that the
principal component will be selected from each of the The approximate weight of every window blocks
window blocks. is calculated by the following formula as given below.
For the ith window blocks
The modified PCA Algorithm for image fusion is
discussed as below: W (Ai) = xi/( xi + yi) and W (Bi) = yi/( xi + yi);
Step 1: Creation Window block for the images. here Ai and Bi represents the ith window block of
the two images and (xi, yi) are the corresponding
Each of the images is split into n window blocks
principal Eigen vectors.
and the number of blocks for both images must be
same. Step 6: Summing two corresponding window
block of images.
JOURNAL OF COMPUTING, VOLUME 3, ISSUE 4, APRIL 2011, ISSN 2151‐9617
HTTPS://SITES.GOOGLE.COM/SITE/JOURNALOFCOMPUTING/
WWW.JOURNALOFCOMPUTING.ORG 172
Adding of the approximation weight of two Correlation Coefficient: It measures the degree of
corresponding window blocks and generating a new correlation between the fused and the reference
fused window block. i.e. ith fused window block images.
Cross Entropy: It reflect the information difference
F= A*W (Ai) + B*W (Bi). between the fused and the reference image.
Entropy: It measures the richness of information in
Step 7: Aggregation of all fused window blocks. the fused image.
Mutual Information: It measures the information
Arranging all the fused window blocks and
shared between the fused image and the reference
getting the final fused image.
image using histograms.
Signal to noise ratio: It measures the ratio between
Finding the covariance matrix for window blocks. information and noise of fused image.
Warping Degree: It measures the level of optical
Determining the eigenvector and then determine principal spectral distortion.
contrast.
JOURNAL OF COMPUTING, VOLUME 3, ISSUE 4, APRIL 2011, ISSN 2151‐9617
HTTPS://SITES.GOOGLE.COM/SITE/JOURNALOFCOMPUTING/
WWW.JOURNALOFCOMPUTING.ORG 173
Here x and y are the pixel values, and are
the contrast of X and Y images. This quality is
measured by breaking the image into 8×8 pixel.
We have also compared the Peak Signal Noise
Ratio and Mutual Information of the fused
image. The results are also compared with the
wavelet based fusion results.
5. EXPERIMENTAL RESULTS AND
ANALYSIS Figure 4: Fused Image by Modified PCA
Table 1: Entropy Measure Comparison
Image Entropy
Input Image A 0.54
Input Image B 0.56
Fused 0.94
Figure 2: Input Image A
Figure 3: Input Image B
JOURNAL OF COMPUTING, VOLUME 3, ISSUE 4, APRIL 2011, ISSN 2151‐9617
HTTPS://SITES.GOOGLE.COM/SITE/JOURNALOFCOMPUTING/
WWW.JOURNALOFCOMPUTING.ORG 174
Table 2: Comparison of the quality parameters
between our modified PCA approach and traditional
PCA
Correlation 1 1
Coefficient
Cross Entropy 0.42 0.47 Figure 6: Comparison chart of Quality of PCA
based and Revised PCA based image.
Entropy 0.88 0.94
From the experimental results as shown in Table
Mutual 0.74 0.71
2 it can be observed that the values of entropy,
Information
mutual information and wrapping degree of the
Mean Square 0.88 0.84 fused image generated by our modified PCA
Error algorithm are greater than values for the fused image
generated by the traditional PCA algorithm. The
Normalized 0.38 0.35 error parameters like the Mean Square Error,
Least square Error Normalized Least square Error and Relative Sift in
Mean have lesser values for our algorithm than the
Relative Sift 0.56 0.52
traditional PCA algorithm. These results clearly show
in Mean
that our modified PCA based image fusion produces
better result than traditional PCA.
Standard 0.001 0.001
Deviation
6. COMPARISION WITH WAVELET
Warping 0.35 0.37 TRANSFORM
degree One of the traditional image fusion techniques is
Wavelet Transform. Popular wavelet based approach
Contrast 0.87 0.88 is to find the decomposition coefficient for image
fusion. The wavelet based method is available as
Luminance 0.91 0.92
image fusion tool in wavelet toolbox which is used
for fusing various registered images of the same size.
The principal of image fusion using wavelet is to
merge the wavelet decompositions of two original
images using fusion methods.
Figure 10: Fused image by traditional PCA
Figure 7: Input image C
Table 3: Quality comparison of PCA based fused
image and our modified PCA based fused image and
the Wavelet Transform
Quality Modified PCA Wavelet
Metric PCA Transform
Figure 8: Input image D
Q 0.79 0.78 0.87
It is clear from Table 3 that our modified PCA is
better than the traditional PCA based results and
very close in quality as compared to the wavelet
based technique. The wavelet based technique has
more complexity than our technique so it can be
inferred that our modified PCA based technique can
Figure 9: Fused image by our algorithm
be useful technique in terms of good quality as well
as reduced complexity.
JOURNAL OF COMPUTING, VOLUME 3, ISSUE 4, APRIL 2011, ISSN 2151‐9617
HTTPS://SITES.GOOGLE.COM/SITE/JOURNALOFCOMPUTING/
WWW.JOURNALOFCOMPUTING.ORG 176
[8] Y. Zheng ‐ “Pyramid, DWT and Iterative DWT ‐ Multi‐scale
7. CONCLUSION
Fusion Algorithm Comparisons.” 12th international conference
This paper presents an algorithm on image on image fusion , 2009 pp 1260 –1267.
fusion which shows much better performance than
[9] S. G. Nikolov, D. R. Bull, C. N. Canagarajah, M. Halliwell, P.
the traditional PCA. The proposed algorithm is based
N. T. Wells ‐ “Image fusion using a 3‐d wavelet transform.”
on statistical measure techniques. We have also
Image Processing and its Applications, Conference Publication
compared the quality of our technique with that of No. 465 0 IEE 1999 pp 235‐239.
traditional PCA based fusion and wavelet based
fusion. From the experimental results it is observed [10] L. I Smit – “A tutorial on Principal Component Analysis”,
2002, pp 1 ‐27.
that the result image have better qualities in terms of
information content as well as lower values of error Mr. Amit Kumar Sen is at present an Assistant Professor in the
measure compared to the traditional PCA. In future Information Technology Depertment, IMPS College of Engg. &
we would like to investigate our algorithm for video Technology, Malda, India. He has done B.Tech in Information
fusion applications. Technology from Bengal Institute of Technology, India (2002‐
2006) and M.Tech in Multimedia and Software System from
NITTTR, Kolkata ( 2006‐2008). His present research area is
Computer Vision and Image Processing.
REFERENCES
Mr. Subhadip Mukherjee is at present working with
[1] Y. Zheng, X. Hou, T. Bian, Z. Qin ‐ Effective Image Fusion CMC.Ltd. , India. Prior to this he was an Assistant Professor in
5th International Symposium on image and signal Processing Engg. & Technology, Malda, India.He is done B.Tech in
[3]S. Li, Z. Li, J. Gong – “Multivariate statistical analysis of Dr. Amlan Chakrabarti is at present an Assistant Professor
measures for assessing the quality of image fusion . (Reader) , in the A.K.Choudhury School of Information
International Journal of Image and Data Fusion”, Volume 1, Technology, University of Calcutta, India. Prior to this he was a
Issue 1 March 2010 , pp 47 – 66. faculty in the Department of Computer Science and
Engineering, West Bengal University of Technology, Meghnad
[4] N. Cvejic, A. Łoza, D. Bull, and N. Canagarajah – “A Novel Saha Institute of Technology, Kolkata and IIIT Calcutta. He is
Metric for Performance Evaluation of Image Fusion an M.Tech. from the University of Calcutta (2001) and has done
Algorithms” World Academy of Science, Engineering and his Doctoral research on Quantum Computing at Indian
Technology 7 2005 pp 80 ‐85. Statistical Institute, India Kolkata, 2004‐2008. He was also a
VLSI Design Engineer from 1998‐2000. He is a Fellow of
[5] Y. Zhu, P. K. Varshney and H.Chen – “Evaluation of ICA Association of Computer Electronics and Electrical Engineers
Based Fusion of Hyperspectral Images for Color Display” . (ACEEE), Senior member of the International Association of
World Academy of Science, Engineering and Technology 7 2005 Computer Science and Information Technology (IACSIT),
Singapore. His present research interests are Quantum
[6] A. Haq, A. M. Mirza and S. Qamar – “An Optimized Image
Computing, VLSI design, Embedded System Design and Video
Fusion Algorithm for Night‐time Surveillance and Navigation”
and Image Processing Algorithms.
Proceedings of IEEE symposium , sept2005 , pp 138 – 143
[7]M. F. Yakhdani and A. Azizi – “Quality assessment of image
fusion techniques for multisensory High resolution satellite
images” (case study: irs‐p5 and irs‐p6 Satellite images) .