1
SVD BASED IMAGE FUSION
SREELEKSHMY SELVIN
M.TECH CEN
P2.CEN.15021
SREELEKSHMY SELVIN
M.TECH CEN
AMRITA SCHOOL OF ENGINEERING
ABSTRACT
 The objective of this work is to demonstrate
the application of Singular Value
Decomposition for fusing images. Image
fusion yields an output image which is
obtained as a combination of two or more
low quality input images. The output image
will have more resolution and is more
informative. Variations in data values are
best analyzed by using SVD; hence in this
work we make use of SVD to extract the
best pixel intensities.
SREELEKSHMY SELVIN
M.TECH CEN
AMRITA SCHOOL OF ENGINEERING 2
INTRODUCTION
 Image fusion : process of combining two
or images with same background.
 It has got applications in the areas like
remote sensing, medical image analysis.
 Can be performed in various levels such
as
pixel level
signal level
feature level
decision level
SREELEKSHMY SELVIN
M.TECH CEN
AMRITA SCHOOL OF ENGINEERING 3
CONT..
 SVD make use of pixel level image fusion
 The various image fusion techniques
include
1) High pass filtering technique
2) IHS transform based image fusion
3) PCA based image fusion
4) Wavelet transforms image fusion
5) Pair-wise spatial frequency matching
SREELEKSHMY SELVIN
M.TECH CEN
AMRITA SCHOOL OF ENGINEERING 4
ALGORITM
 READ INPUT IMAGES
 READ THE REFERENCE IMAGE
 CONVERT ALL IMAGES TO GRAYSCALE
 FIND SVD OF INPUT IMAGES
 FIND THE PROMINENT VALUES FROM
THE S VALUES OF BOTH IMAGES
SREELEKSHMY SELVIN
M.TECH CEN
AMRITA SCHOOL OF ENGINEERING 5
CONT…
 FIX THE PARAMETERS K,W
 EXTRACT THE 'K' PROMINENT VALUES
FROM THE DECOMPOSED MATRICES
 FUSE THE IMAGE AND APPLY
WEIGHING FACTOR
 DISPLAY THE OUTPUT IMAGE AND
REFERENCE IMAGE
 FIND THE PERFORMANCE FACTORS
SREELEKSHMY SELVIN M.TECH CEN AMRITA
SCHOOL OF ENGINEERING 6
INPUT IMAGES
SREELEKSHMY SELVIN
M.TECH CEN
AMRITA SCHOOL OF ENGINEERING 7
OUTPUT IMAGE &
REFERENCE IMAGE
SREELEKSHMY SELVIN
M.TECH CEN
AMRITA SCHOOL OF ENGINEERING 8
OUTPUT IMAGE REFERENCE IMAGE
PERFORMANCE
FACTORS
 The performance factors gives the
measure of similarity between the output
image and the reference image
 Mean Square Error (MSE)
 average of squares of error
 image quality increases as MSE
decreases
SREELEKSHMY SELVIN
M.TECH CEN
AMRITA SCHOOL OF ENGINEERING 9
CONT…
 PEAK SIGNAL TO NOISE RATIO
(PSNR)
 ratio between the maximum possible power of
a signal and the power of noise that affects
the output.
 image quality increases as PSNR increases.
SREELEKSHMY SELVIN
M.TECH CEN
AMRITA SCHOOL OF ENGINEERING 10
CONT..
NORMALISE
CROSSCORRELATION
(NC)
 measure of similarity of
two images.
 image quality increases
as NC increases.
 STRUCTURED
CONTENT (SC)
 ratio between the content of
both the expected and the
obtained data
 Practically it is the square of
the expected data and the
net sum of square of the
obtained data.
SREELEKSHMY SELVIN
M.TECH CEN
AMRITA SCHOOL OF ENGINEERING 11
CONT…
 NORMALISED
ABSOLUTE ERROR
(NAE)
 the net sum ratio
between the error
values and the
perfect values is
calculated.
SREELEKSHMY SELVIN
M.TECH CEN
AMRITA SCHOOL OF ENGINEERING 12
RESULT
 SVD based image fusion retrieves the
relevant information from the singular
values and performs fusion.
 It is more effective than the other
existing methods.
SREELEKSHMY SELVIN
M.TECH CEN
AMRITA SCHOOL OF ENGINEERING 13
SREELEKSHMY SELVIN M.TECH CEN AMRITA
SCHOOL OF ENGINEERING 14

SVD

  • 1.
    1 SVD BASED IMAGEFUSION SREELEKSHMY SELVIN M.TECH CEN P2.CEN.15021 SREELEKSHMY SELVIN M.TECH CEN AMRITA SCHOOL OF ENGINEERING
  • 2.
    ABSTRACT  The objectiveof this work is to demonstrate the application of Singular Value Decomposition for fusing images. Image fusion yields an output image which is obtained as a combination of two or more low quality input images. The output image will have more resolution and is more informative. Variations in data values are best analyzed by using SVD; hence in this work we make use of SVD to extract the best pixel intensities. SREELEKSHMY SELVIN M.TECH CEN AMRITA SCHOOL OF ENGINEERING 2
  • 3.
    INTRODUCTION  Image fusion: process of combining two or images with same background.  It has got applications in the areas like remote sensing, medical image analysis.  Can be performed in various levels such as pixel level signal level feature level decision level SREELEKSHMY SELVIN M.TECH CEN AMRITA SCHOOL OF ENGINEERING 3
  • 4.
    CONT..  SVD makeuse of pixel level image fusion  The various image fusion techniques include 1) High pass filtering technique 2) IHS transform based image fusion 3) PCA based image fusion 4) Wavelet transforms image fusion 5) Pair-wise spatial frequency matching SREELEKSHMY SELVIN M.TECH CEN AMRITA SCHOOL OF ENGINEERING 4
  • 5.
    ALGORITM  READ INPUTIMAGES  READ THE REFERENCE IMAGE  CONVERT ALL IMAGES TO GRAYSCALE  FIND SVD OF INPUT IMAGES  FIND THE PROMINENT VALUES FROM THE S VALUES OF BOTH IMAGES SREELEKSHMY SELVIN M.TECH CEN AMRITA SCHOOL OF ENGINEERING 5
  • 6.
    CONT…  FIX THEPARAMETERS K,W  EXTRACT THE 'K' PROMINENT VALUES FROM THE DECOMPOSED MATRICES  FUSE THE IMAGE AND APPLY WEIGHING FACTOR  DISPLAY THE OUTPUT IMAGE AND REFERENCE IMAGE  FIND THE PERFORMANCE FACTORS SREELEKSHMY SELVIN M.TECH CEN AMRITA SCHOOL OF ENGINEERING 6
  • 7.
    INPUT IMAGES SREELEKSHMY SELVIN M.TECHCEN AMRITA SCHOOL OF ENGINEERING 7
  • 8.
    OUTPUT IMAGE & REFERENCEIMAGE SREELEKSHMY SELVIN M.TECH CEN AMRITA SCHOOL OF ENGINEERING 8 OUTPUT IMAGE REFERENCE IMAGE
  • 9.
    PERFORMANCE FACTORS  The performancefactors gives the measure of similarity between the output image and the reference image  Mean Square Error (MSE)  average of squares of error  image quality increases as MSE decreases SREELEKSHMY SELVIN M.TECH CEN AMRITA SCHOOL OF ENGINEERING 9
  • 10.
    CONT…  PEAK SIGNALTO NOISE RATIO (PSNR)  ratio between the maximum possible power of a signal and the power of noise that affects the output.  image quality increases as PSNR increases. SREELEKSHMY SELVIN M.TECH CEN AMRITA SCHOOL OF ENGINEERING 10
  • 11.
    CONT.. NORMALISE CROSSCORRELATION (NC)  measure ofsimilarity of two images.  image quality increases as NC increases.  STRUCTURED CONTENT (SC)  ratio between the content of both the expected and the obtained data  Practically it is the square of the expected data and the net sum of square of the obtained data. SREELEKSHMY SELVIN M.TECH CEN AMRITA SCHOOL OF ENGINEERING 11
  • 12.
    CONT…  NORMALISED ABSOLUTE ERROR (NAE) the net sum ratio between the error values and the perfect values is calculated. SREELEKSHMY SELVIN M.TECH CEN AMRITA SCHOOL OF ENGINEERING 12
  • 13.
    RESULT  SVD basedimage fusion retrieves the relevant information from the singular values and performs fusion.  It is more effective than the other existing methods. SREELEKSHMY SELVIN M.TECH CEN AMRITA SCHOOL OF ENGINEERING 13
  • 14.
    SREELEKSHMY SELVIN M.TECHCEN AMRITA SCHOOL OF ENGINEERING 14