Professional Documents
Culture Documents
CHAPTER-1
INTRODUCTION
1.1 Image
In common usage, an image or picture is an artifact that produces the likeness of
some subject–usually a physical object or a person. Images may be two dimensional (e.g.
a photograph) or three dimensional (e.g. a statue). They are typically produced by optical
devices–such as cameras, mirrors, lenses, telescopes, microscopes, etc. and natural
objects and phenomena, such as the human eye or water surfaces. The word image is also
used in the broader sense of any two dimensional figures or illustration, e.g. a map, a
graph, a pie chart, an abstract painting, etc. In this wider sense, images can also be
produced manually (by drawing painting, carving, etc.), by computer graphics
technology, or a combination of the two.
Digital Image
A digital image is a representation of a two–dimensional image as a finite set of
digital values, called picture elements or pixels. Typically, the pixels are stored in
computer memory as a raster image or raster map, a two–dimensional array of small
integers. These values are often transmitted or stored in a compressed form. Digital
images can be created by a variety of input devices and techniques, such as digital
cameras, scanners, coordinate–measuring machines, seismographic profiling, airborne
radar, and more.
1.1.1 Pixel
A pixel is one of the many tiny dots that make up the representation of a picture in
a computer„s memory. Each such information element is not really a dot, nor a square,
but an abstract sample. With care, pixels in an image can be reproduced at any size
without the appearance of visible dots or squares; but in many contexts, they are
reproduced as dots or squares and can be visibly distinct when not fine enough.
The intensity of each pixel is variable; in color systems, each pixel has typically
three or four dimensions of variability such and Red, Green and Blue, or Cyan, Magenta,
Yellow and Black.
Sub Pixels
Many display and image-acquisition systems are, for various reasons, not capable
of displaying the different color channels at the same site. This approach is generally
resolved by using multiple sub pixels, each of which handles a single color channel. For
example, LCD displays typically divide each pixel into four sub pixels; one red, one
green, and two blue. Most digital camera sensors also use sub pixels by using colored
filters.
For systems with sub pixels two different approaches can be taken: the sub pixels
can be ignored with pixels being treated as the smallest addressable imaging element, or
the sub pixels can be included in rendering calculations, which requires more analysis
and processing time, but can produce apparently superior images in some cases. The later
approach has been used to increase the apparent resolution of color displays.
Mega Pixel
A mega pixel is 1 million pixels, and is usually used to express the resolution
capabilities of digital cameras. For example, a camera that can take pictures with a
resolution of 2048 x 1536 pixels is commonly said to have 3.1 mega pixels (2048 x
1536=3,145,728). Digital cameras use photo sensitive electronics; either Charge-coupled
devices (CCDs) or CMOS sensors, which record brightness levels on a per-pixel basis.
1.1.2 Images in Matlab
The basic data structure in MATLAB is the array, an ordered set of real or
complex elements. This object is naturally suited to the representation of images, real-
valued ordered sets of color or intensity data
MATLAB stores most images as two-dimensional arrays (i.e., matrices), in which
each element of the matrix corresponds to a single pixel in the displayed image. (Pixel is
derived from picture element and usually denotes a single dot on a computer display.)
1.1.3 Image Representation
An image is stored as a matrix using standard MATLAB matrix conventions.
There are four basic types of images supported by MATLAB.
1. Binary images
2. Intensity images
3. RGB images
4. Indexed images
1. Binary Image
In a binary image, each pixel assumes one of only two discrete values: 1 or 0. A
binary image is stored as a logical array. By convention, this documentation uses the
variable name BW to refer to binary images.
The following figure shows a binary image with a close-up view of some of the
pixel values.
2. Grayscale Image
A gray scale image (also called gray-scale or gray-level) is a data matrix whose
values represent intensities within some range. MATLAB stores a grayscale image as an
individual matrix, with each element of the matrix corresponding to one image pixel. By
convention, this documentation uses the variable name I to refer to grayscale images.
The matrix can be of class uint8, uint16, int16, single, or double. While grayscale
images are rarely saved with a color map, MATLAB uses a colour map to display them.
For a matrix of class single or double, using the default gray scale colour map, the
intensity 0 represents black and the intensity 1 represents white. For a matrix of type
uint8, uint16, or int16, the intensity intmin (class (I)) represents black and the intensity
intmax (class (I)) represents white. The figure below depicts a grayscale image of class
double.
Fig 1.2: Pixel Values in a Gray scale, Image Define Gray Levels
3. RGB Image
A color image is an image in which each pixel is specified by three values one
each for the red, blue, and green components of the pixel's color. MATLAB store color
images as an m-by-n-by-3 data array that defines red, green, and blue color components
for each individual pixel. Color images do not use a color map. The color of each pixel is
determined by the combination of the red, green, and blue intensities stored in each color
plane at the pixel's location.
A color array can be of class uint8, uint16, single, or double. In a color array of
class single or double, each color component is a value between 0 and 1. A pixel whose
color components are (0, 0, 0) is displayed as black, and a pixel whose color components
are (1, 1, 1) is displayed as white. The three color components for each pixel are stored
along the third dimension of the data array. For example, the red, green, and blue color
components of the pixel (10,5) are stored in RGB(10,5,1), RGB(10,5,2), and
RGB(10,5,3), respectively The following figure depicts a color image of class double.
4. Indexed Image
An indexed image consists of an array and a colour map matrix. The pixel values
in the array are direct indices into a color map. By convention, this documentation uses
the variable name X to refer to the array and map to refer to the color map.
The relationship between the values in the image matrix and the color map
depends on class of the image matrix. If the image matrix is of class single or double, it
normally contains integer values 1 through p, where p is the length of the color map. The
value 1 points to the first row in the color map, the value 2 points to the second row, and
so on. If the image matrix is of class logical, uint8 or uint16, the value 0 points to the first
row in the color map, the value 1 points to the second row, and so on.
The following figure illustrates the structure of an indexed image. In the figure,
the image matrix is of class double, so the value 5 points to the fifth row of the color
map.
Fig 1.4: Pixel Values Index to Color map Entries in Indexed Image
digital images that you will want to print. Tiff supports embedded file information,
including exact color space, output profile information and EXIF data. There is a lossless
compression for TIFF called LZW. LZW is much like 'zipping' the image file because
there is no quality loss. An LZW TIFF decompresses (opens) with all of the original pixel
information unaltered.
4. BMP (Windows Bitmap)
It is a Windows (only) operating system uncompressed file format that supports
24 bit color. BMP does not support embedded information like EXIF, calibrated colour
space and output profiles. Avoid using BMP for photographs because it produces
approximately the same file sizes as TIFF without any of the advantages of TIFF.
5. Camera Raw
It is a lossless compressed file format that is proprietary for each digital camera
manufacturer and model. A camera RAW file contains the 'raw' data from the camera's
imaging sensor. Some image editing programs have their own version of RAW too.
However, camera RAW is the most common type of RAW file. The advantage of camera
RAW is that it contains the full range of colour information from the sensor. This means
the RAW file contains 12 to 14 bits of colour information for each pixel.
If you shoot JPEG, you only get 8 bits of colour for each pixel. These extra colour bits
make shooting camera RAW much like shooting negative film. You have a little more
latitude in setting your exposure and a slightly wider dynamic range.
1.1.5 Image Coordinate Systems
Pixel Coordinates
Generally, the most convenient method for expressing locations in an image is to
use pixel coordinates. In this coordinate system, the image is treated as a grid of discrete
elements, ordered from top to bottom and left to right, as illustrated by the following
figure.
Unit16 Unsigned 16-bit integers in the range [0, 65535] (2 bytes per
element).
Int8 Signed 8-bit integers in the range [-128, 127] (1 byte per
element).
Int16 Signed 16-bit integers in the range [-32768, 32767] (2 bytes per
element)
Certain MATLAB functions, including the find, all, any, conv2, convn, fft2, fftn,
and sum functions, accept uint8 or uint16 data but return data in double-precision format.
Function Description
Color processing.
Pattern recognition.
Video processing
1.4 Motivation
Image fusion is a technique of fusing multiple Images for better information and
more accurate image compared to Input images. Different types of image fusion process
are used to increase the visibility of an image. Depend upon the requirement image fusion
types will be varied. The proposed methodology uses benefits of NSCT and SVD to fuse
the two images. The Non sub-sampled contourlet (NSCT) type of fusion method have
more advantages like multi scale, multidirectional expansion and provides better
frequency selectivity. The Singular Value Decomposition (SVD) based image processing
techniques were focused in compression, watermarking and quality measure.
1.5 Need of Image Fusion
Fusion technique can be applied for various methods. When compared with other
methods, NSCT (Non-sub sampled contourlet transform) provides quality to the image at
the edges by removing the noise, and analyzes feature of the fused images better. NSCT
provides better frequency selectivity and regularity when compared to the CT (contourlet
transform).Using only NSCT or only SVD makes the algorithm complex so we use
combination of NSCT and SVD.
The SVD is the optimal matrix decomposition in a least square sense that it packs
the maximum signal energy into as few coefficients as possible. The major advantages of
SVD are image compression, watermarking and quality measure. Among all the
techniques, NSCT and SVD are the major important transformation technique for the
fusion process of an image.
1.6 Objectives and Goals
Image fusion is a technology that keeps that images are the main research
contents and it refers the techniques that to integrate the multi images of the same scene
or multi images of one sensor. The main goal of this paper presents that to search the
algorithms and that can be used to implement the image fusion in various applications
like runway extraction etc. To evaluate the performance with different image quality
terms and those properties are chosen with great impact factor to detect the image fusion
algorithms.
To obtain an informative image using NSCT and SVD, the major advantage is
NSCT is a kind of multi-scale and multi-direction computation framework of the discrete
images which can be divided into two stages includes Non-Sub sampled Pyramid (NSP)
and Non-Sub sampled Directional filter bank (NSDFB) to improve the quality of the
image.SVD reduces the noise of the image. The major goal is that implement that fusion
process by using the NSCT and SVD techniques.
1.7 Software Requirement
MATLAB, Version: R2017a, 64-bit
1.8 Organization of Thesis
In chapter-1 it explains Introduction to digital Image processing and different types of
images and its applications.
In chapter-2 it explains literature survey, different methods in Image fusion like DCT,
SVD, NSCT, SWT etc, and also includes existing methods and its drawbacks.
In chapter-3 it explains about the definition of Image fusion its techniques, steps, and its
applications in different domains.
In chapter-4 the proposed algorithm, the major aim of the project is explained.
In chapter-5 it introduces the MATLAB and its working environment.
In Chapter-6 it discusses the performance characteristics of NSCT& SVD based Image
fusion technique and its results.
In Chapter-7 it concludes the thesis and future scope of project further research.
CHAPTER-2
LITERATURE SURVEY
2.1 Introduction
At present, the resolution of image is one of the major issues in functioning at
under low visibility conditions. So, various kinds of issues are proposed for increasing
the visibility conditions. Among all the techniques, NSCT and SVD are the major
important transformation technique for the fusion process of an image.
2.2 Research Papers
In the part of literature survey first consider the research papers for project to
implement the image fusion technique for vision systems is as follows
Research Paper-1 (1)
Maes, D. Vandermeulen, and P. Suetens propose - Analysis of multispectral or
multitemporal images requires proper geometric alignment of the images to compare
corresponding regions in each image volume. Retrospective three-dimensional alignment
or registration of multimodal medical images based on features intrinsic to the image data
itself is complicated by their different photometric properties, by the complexity of the
anatomical objects in the scene and by the large variety of clinical applications in which
registration is involved. While the accuracy of registration approaches based on matching
of anatomical landmarks or object surfaces suffers from segmentation errors, voxel-based
approaches consider all voxels in the image without the need for segmentation. The
recent introduction of the criterion of maximization of mutual information, a basic
concept from information theory, has proven to be a breakthrough in the field. While
solutions for intrapatient affine registration based on this concept are already
commercially available, current research in the field focuses on interpatientnonrigid
matching.
based on features intrinsic to the image data itself is complicated by their different
photometric properties, by the complexity of the anatomical objects in the scene and by
the large variety of clinical applications in which registration is involved. While the
accuracy of registration approaches based on matching of anatomical landmarks or object
surfaces suffers from segmentation errors, voxel-based approaches consider all voxels in
the image without the need for segmentation. The recent introduction of the criterion of
maximization of mutual information, a basic concept from information theory, has
proven to be a breakthrough in the field. While solutions for intrapatient affine
registration based on this concept are already commercially available, current research in
the field focuses on interpatientnonrigid matching.
Research Paper-2 (2)
A. Cardinali and G. P. Nason propose an algorithm to adaptively segment and
fuse images by alternating wavelet packet and local cosine transforms each containing
best basis selection and thresholding. Within segmented regions fusion is informed by
multiple hypothesis testing based on a log-linear factorial model. This fusion identifies
homogenous regions from which to select wavelet or local cosine packets, possibly from
the original images. The successful performance of the fusion algorithm and
segmentation is demonstrated on some multispectral thematic mapper imagery. With the
increasing availability of large amounts of various kinds of data sources, each
characterizing different kinds of phenomena, a need has arise for statistical and
mathematical methods that are capable of capturing complementary information and
merging it in an efficient way. The purposes of merging might be for human presentation
or for further processing with techniques that might not be able to handle the original
data. As an example, we fuse some multispectral images: images of the same scene but
sensed using different frequencies. The example that we show later in Figure 2 depicts
Chew Valley Lake in Somerset, UK and is available at 12 frequencies. The task we
attempt is to fuse two of these bands into one that contains important features available
separately in one of the images but not the other.
Research Paper-3 (3)
Ding Li proposes a new method of fusing panchromatic and multispectral images
based on NSCT and PCA. PCA reduces the dimensionality but preserves maximum
possible information of the data sources. It transforms a vector of multivariate data with
correlated variables into uncorrelated variables. Considering the advantages of NSCT,
this paper proposes a new method of fusing remote sensing image combining NSCT with
PCA. The final experimental results show that it has better subjective visual effect and
objective evaluation compared to other methods. First we perform PCA on the
multispectral image, and get principal components. Then apply histogram matching
between the original panchromatic image and PC1 to get approximate mean value. Now
employ NSCT on PC1 and panchromatic image to get low frequent and high frequent sub
bands. Fuse the PC1 and panchromatic image and apply NSCT reconstruction with new
coefficient to obtain the new PC1.Finally perform the inverse PCA transform to obtain
the fused image.
Research Paper-4 (4)
Gaurav Bhatnagar, Q. M. Jonathan Wu, Balasubramanian Raman proposes
Image Fusion is a technique which attempts to combine complimentary information
from multiple images of the same scene so that the fused image is more suitable for
computer processing tasks and human visual system. In this paper, a simple yet
efficient real time image fusion algorithm is proposed considering human visual
properties in spatial domain. The algorithm is computationally simple and implemented
very easily in real-time applications. Experimental results highlights the expediency
and suitability of the algorithm and efficiency is carried by the comparison made
between proposed and existing algorithm.
Research Paper-5 (5)
Kurakula Sravya, Dr. P. Govardhan, Naresh Goud M proposes a paper on
Image Fusion on Multi Focused Images Using NSCT. Image fusion is the process that
combines information in multiple images of the same scene. These images may be
captured from different sensors, acquired at different times, or having different spatial
and spectral characteristics. The object of the image fusion is to retain the most desirable
characteristics of each image. With the availability of multi-sensor data in many fields,
image fusion has been receiving increasing attention in the researches for a wide
b. Pyramid method:
(i) Gaussian pyramid
(ii) Laplacian Pyramid
(iii) Gradient pyramid
(iv) Morphological pyramid
(v) Ratio of low pass pyramid
c. Wavelet transforms:
(i) Discrete wavelet transforms (DWT)
(ii) Stationary wavelet transforms
(iii) Multi-wavelet transforms
d. Curvelet transforms
2.3.1 IHS Transform
IHS transform is the oldest method for image fusion process. Intensity, Hue and
saturation are the three properties of colours and give a visibility perception of image.
Hue and Saturation contains more spectral information compare to intensity of image.
This method gives a fused output but not recent technique.
2.3.2 Principal Component Analysis (PCA)
The advantage of PCA when compared with IHS is it uses the arbitrary number of
bands. It is one of the important methods to perform the fusion process. Uncorrelated
principal components are formed from the low resolution multi spectral images. The first
PCA1 component is having the information of the variance and it gives the more
effective information of panchromatic image. Then an inverse PCA is used for the fused
image.
PCA is a mathematical tool transforms correlated variables into uncorrelated
variables called principal components. It is used in mainly image classification and image
compression. The first principal component occurs for the maximum variance in an
image and second principal component occurs is at subspace perpendicular to the first
component. The third principal component is subspace perpendicular to the first and two
and so on. This is the way of occuring the principal components in PCA method.
The image fusion process by using Principal Component Analysis (PCA) is as follows in
FIG 2.3:
Image I1(x, y)
Image I2(x, y)
and the low pass filtered image. Laplacian pyramid is a set of band pass filters it can be
used to represent images as a series of band pass filtered images, each sampled at
successively sparser density. It is frequently used in image processing and pattern
recognition.
Ratio of Low Pass Pyramid
In ratio of low pass pyramid we take the ratio of two successive layers. The ROLP
pyramid is a complete representation of the original image. A ROLP pyramid is
constructed for each of the source images .As it is constructed for the composite image
from the corresponding nodes in the component pyramids, the one with maximum
absolute contrast.
Morphological Pyramid
Morphological pyramids systematically split the input signal in to approximation
and detail signals by repeatedly applying morphological filters follow by down sample.
The fundamental morphological operators are: erosion, dilation, opening and closing.
Consistent analysis of techniques will help in deciding the suitability of a particular
technique towards the fusion of large number of images.
2.3.4 Wavelet transforms
Discrete Cosine Transform (DCT)
DCT comprises the images in the form of MPEG, JVG etc. In this transform
technique the spatial domain image will be converted into frequency domain image. The
2 dimensional DCT will be applied on gray scale image and the frequency of gray scale
image will be converted spatial domain into frequency domain. Fused DCT coefficients
are obtained by the fusion rule. By using the inverse DCT transform the fused image will
be obtained. It is one of the most spatial domain fusion methods. These methods are
complex and time consuming process. These are hard to perform. When real time
applications, the source images are coded in JPEG and in JPEG format, the fusion
approaches are applied in DCT domain is very efficient. The DCT operation is performed
on each and every block and it generates 64 coefficients to reduce the magnitude.
Those coefficients are rearranged in nonlinear manner for their further encoding
process. In case of using spatial domain the images are decoded and transferred and then
after applying fusion procedure the fused image will be coded again.
Stationary Wavelet Transform (SWT)
The stationary wavelet transform (SWT) is an expansion of standard discrete
wavelet transform (DWT) that utilizes high and low pass channels. SWT apply high and
low pass channels to the information at every level and at next stage it produces two
sequences. The two new successions will have same length as that of first grouping. In
SWT, rather than annihilation the channels at every level is altered by cushioning them
with zeroes. Stationary Wavelet Transform is computationally more complex. The
Discrete Wavelet Transform is a time variant transform. The best approach to restore the
interpretation invariance is to average some slightly distinctive DWT, called undecimated
DWT to characterize the stationary wavelet transform (SWT). SWT does this by
suppressing the down-sampling step of the DWT and instead up-sampling the filters by
padding with zeros between the filter coefficients. After decomposition, four images are
generally furnished (one approximation and three detail coefficients) which are at half the
resolution of the original image in DWT, whereas in SWT the approximation and detail
coefficients will have the same size as the input images. SWT is like discrete wavelet
transforms (DWT), however the main procedure of down-sampling is stifled which
implies that SWT is shift invariant. It applies the DWT and excludes both down-
sampling in the forward and up-sampling in the reverse direction. More precisely, it
executes the transform at each point of the image and saves the detail coefficients and
uses the low frequency information at each level.
principle highlight of DWT is multi scale representation. By utilizing the wavelets, given
functions can be analyzed at different levels of resolution. DWT decomposition utilizes a
course of low pass and high-pass channels and a sub-sampling operation. The yields from
2D-DWT are four images having size equal to half the size of input image. So from first
input image HHa, HLa, LHa, LLa images are obtained and from second input image
HHb, HLb, LHb, LLb images are obtained. Here LL image contains the approximation
coefficients. LH image contains the horizontal detail coefficients. HL image contains the
vertical detail coefficients and HH contains the diagonal detail coefficients. One of the
significant disadvantages of wavelet transform is their absence of translation invariance.
Curvelet Transform
In addition to shift-invariance, it has been recognized that an efficient image
representation has to account for the geometrical structure pervasive in natural scenes. In
this direction the contourlet transform is a multidirectional and multi scale transform that
is constructed by combining the Laplacian pyramid with the directional filter bank (DFB)
proposed in curvelet transform, which represents edges better than wavelets .The
pyramidal filter bank structure of the contourlet transform has very little redundancy,
which is important for compression applications. However, designing good filters for the
contourlet transform is a difficult task. In addition, due to down samplers and up samplers
present in both the Laplacian pyramid and the DFB, the contourlet transform is not shift-
invariant.
CHAPTER-3
IMAGE FUSION
Introduction
Image fusion is the process of combining two or more images into a single image.
The resulting image will be more informative than source images. These images may be
captured from different sensors, acquired at different times or having different spatial and
spectral characteristics. With the availability of multi sensor data in many fields, image
fusion has been receiving increasing attention in the researches for a wide spectrum of
applications. Image fusion has been widely used in military, remote sensing, robot vision,
medical image processing and other areas.
3.1 Definition of Image Fusion
Image fusion is type of data fusion, which can be defined as the process of
combining two or more source images from the same scene into a composite image with
extend information content by using a certain algorithm. The fused image may provide
increased interpretation capabilities and more reliable results since data with different
characteristics. Moreover, image fusion can be performed at three different processing
levels according to the stage at which the fusion takes: pixel, feature and decision level.
Split the images into two or more up to n depend upon the requirement.
Fusion process is applied by using different techniques like NSCT, PCA, DCA
etc.
Finally the fused image is obtained and it contains all the information, more
effective when compared with the input image.
Input Image
Image Decomposition/Split
Image Fusion
condition. A variety of machines and techniques can create pictures of the structures and
activities inside your body. The type of imaging your doctor uses depends on your
symptoms and the part of your body being examined. Each type of technology gives
different information about the area of the body being studied or treated, related to
possible disease, injury, or the effectiveness of medical treatment. The medical imaging
techniques include
X-rays
MRI scan
CT scan etc.
Many imaging tests are painless and easy. Some require you to stay still for a long
time inside a machine. This can be uncomfortable. Certain tests involve exposure to a
small amount of radiation.
3.5 Image Fusion Categories
Image fusion operation can perform in four levels such as pixel level, signal level,
feature level and decision level.
In Signal Level Fusion, the signal from different sensors is combined to produce
a new signal with better signal to noise ratio than the original signals.
In Pixel Level Fusion, it performs the operation on every pixel and produces the
fused image information from a set of pixels in source images to improve the
performance of image.
In Feature Level Fuision, it requires an extraction of objects from various data
sources. It requires the information from pixel intensities, edges and textures.
In Decision Level Fusion, the information is merging at higher level of
abstraction and it combines the results from multiple algorithms and finally gets
the fused image.
CHAPTER-4
PROPOSED METHODOLOGY
4.1 Introduction
The image fusion method widely uses wavelet transform, contourlet transform
and Non-sub sampled contourlet transform. The wavelet transform can preserve spectral
information efficiently but cannot capture the image geometry structure. Furthermore the
isotropic wavelets are scant of shift invariance and multi directionality and fail to
provide, and optimal expression of highly anisotropic edges and contours in images. The
CT and NSCT overcome the short coming, which have advantages of localization,
directionality and anisotropy.
The NSCT is a fully shift invariance, multi scale and multi direction expansion
whose core is non separable two channel non subsamples filter bank(NSFB).The less
stringent design condition of the NSFB to design filters leads to NSCT with better
frequency selectivity and regularity when compared CT. To achieve shift-invariance the
NSCT is built upon coupling a Non-sub sampled Pyramid (NSP) with the Non-sub
sampled Directional filter bank (NSDFB). The multi scale property of the NSCT is
obtained from NSP which is a two-channel NSFB. NSP is completely different from the
Laplacian Pyramid (LP) in the CT because it has no down samp ling or up sampling,
hence it is shift-invariance.
The salient features of NSCT over existing methodology are as follows:
Two different fusion rules are proposed for combining low and high-frequency
coefficients.
For fusing the low-frequency coefficients, the phase congruency based model is
used. The main benefit of phase congruency is that it selects and combines
contrast- and brightness - invariant representation contained in the low frequency
coefficients.
On the contrary, a new definition of directive contrast in NSCT domain is
proposed and used to combine high-frequency coefficients. Using directive
contrast, the most prominent texture and edge information are selected from high-
frequency coefficients and combined in the fused ones.
The definition of directive contrast is consolidated by incorporating a visual
constant to the SML based definition of directive contrast which provides a richer
representation of the contrast.
Image 1 Image 2
NSCT NSCT
Fusion based on MSVD and applies IMSVD Fusion based on MSVD and applies IMSVD
INSCT
Fused image
channel fan filter banks and re-sampling operation. The result is a tree-structured filter
bank that splits the 2-D frequency plane into directional wedges. A shift-invariant
directional expansion is obtained with a non-sub sampled DFB (NSDFB). The NSDFB is
constructed by eliminating the down-samplers and up samplers in the DFB. This is done
by switching off the down samplers/ up samplers in each two-channel filter bank in the
DFB tree structure and up sampling the filters accordingly. This results in a tree
composed of two-channel NSFBs.
The Non-sub sampled contour let transform (NSCT) is used in the proposed
framework. NSCT has properties such as multistate, localization, multidirectional, and
shift invariance, but only limits the signal analysis to the time frequency domain. Two
different fusion rules are proposed for combining low and high-frequency coefficients.
For fusing the low-frequency coefficients, the phase congruency based model is used. A
new definition of directive contrast in
NSCT domain is proposed and used to combine high-frequency coefficients. Finally, the
fused image is constructed by the inverse NSCT with all composite coefficients.
Low Coefficients A
Fusion
Image A
Rule 1
Low Coefficients A
Fused Fused
NSCT
coeffi Image
Low Coefficients B cients
Fusion
Image B Rule 2
Low Coefficients B INSCT
NSCT
source image.
4.4 SVD (Singular Value Decomposition)
Singular Value Decomposition (SVD) has recently emerged as a new paradigm
for processing different types of images. SVD is an attractive algebraic transform for
image processing applications.
The SVD is the optimal matrix decomposition in a least square sense that it packs
the maximum signal energy into as few coefficients as possible. Singular value
decomposition (SVD) is a stable and effective method to split the system into a set of
linearly independent components, each of them bearing own energy contribution.
Singular value decomposition (SVD) is a numerical technique used to diagonalize
matrices in numerical analysis. SVD is an attractive algebraic transform for image
processing, because of its endless advantages, such as maximum energy packing which is
usually used in compression, ability to manipulate the image in base of two distinctive
subspaces data and noise subspaces, which is usually uses in noise filtering and also was
utilized in watermarking applications.
Each of these applications exploits key properties of the SVD. Also it is usually
used in solving of least squares problem, computing pseudo- inverse of a matrix and
multivariate analysis. SVD is robust and reliable orthogonal matrix decomposition
methods, which is due to its conceptual and stability reasons becoming more and more
popular in signal processing area. SVD has the ability to adapt to the variations in local
statistics of an image. Many SVD properties are attractive and are still not fully utilized.
The developed SVD based image processing techniques were focused in compression,
watermarking and quality measure. Experiments are performed to validate some of well
known but unutilized properties of SVD in image processing applications. This
contributes in utilizing SVD generous properties that are not unexploited in image
processing. It also introduces new trends and challenges in using SVD in image
processing applications. Some of these new trends are well examined experimentally and
validated and others are demonstrated and needs more work to be maturely validated. It
opens many tracks for future work in using SVD as an imperative tool in signal
processing.
0
The singular value decomposition of a matrix A is the factorization of A into the
product of three matrices A = UDVT where the columns of U and V are orthonormal and
the matrix D is diagonal with positive real entries. The SVD is useful in many tasks. Here
we mention some examples. First, in many applications, the data matrix A is close to a
matrix of low rank and it is useful to find a low rank matrix which is a good
approximation to the data matrix. We will show that from the singular value
decomposition of A, we can get the matrix B of rank k which best approximates A; in
fact we can do this for every k. Also, singular value decomposition is defined for all
matrices (rectangular or square) unlike the more commonly used spectral decomposition
in Linear Algebra. The reader familiar with eigenvectors and eigenvalues (we do not
assume familiarity here) will also realize that we need conditions on the matrix to ensure
orthogonality of eigenvectors. In contrast, the columns of V in the singular value
decomposition, called the right singular vectors of A, always form an orthogonal set with
no assumptions on A. The columns of U are called the left singular vectors and they also
form an orthogonal set. A simple consequence of the orthogonality is that for a square
and invertible matrix A, the inverse of A is V D-1 UT, as the reader can verify. To gain
insight into the SVD, treat the rows of an n × d matrix A as n points in a d-dimensional
space and consider the problem of finding the best k- dimensional subspace with respect
to the set of points. Here best means minimize the sum of the squares of the
perpendicular distances of the points to the subspace. We begin with a special case of the
problem where the subspace is 1-dimensional, a line through the origin. We will see later
that the best-fitting k-dimensional subspace can be found by k applications of the best
fitting line algorithm. Finding the best fitting line through the origin with respect to a set
of points {xi |1 ≤ i ≤ n} in the plane means minimizing the sum of the squared distances
of the points to the line. Here distance is measured perpendicular to the line. The problem
is called the best least squares fit. In the best least squares fit, one is minimizing the
distance to a subspace. An alternative problem is to find the function that best fits some
data. Here one variable y is a function of the variables x1, x2, _ _ _, xd and one wishes to
minimize the vertical distance, i.e., distance in the y direction, to the subspace of the xi
rather than minimize the perpendicular distance to the subspace being fit to the data.
4.4.1. Singular Vectors
We now define the singular vectors of an n × d matrix A. Consider the rows of A
as n points in a d-dimensional space. Consider the best fit line through the origin. Let v
be a unit vector along this line. The length of the projection of ai , the i th row of A, onto
v is |ai · v|. From this we see that the sum of length squared of the projections is |Av| 2.
The best fit line is the one maximizing |Av| 2 and hence minimizing the sum of the
squared distances of the points to the line. With this in mind, define the first singular
vector, v1, of A, which is a column vector, as the best fit line through the origin for the n
points in d-space that are the rows of A. Thus v1 = arg max |v|=1 |Av|. The value σ1 (A)
= |Av1| is called the first singular value of A. Note that σ 2 1 is the sum of the squares of
the projections of the points to the line determined by v1. The greedy approach to find the
best fit 2-dimensional subspace for a matrix A, takes v1 as the first basis vector for the 2-
dimenional subspace and finds the best 2-dimensional subspace containing v1. The fact
that we are using the sum of squared distances will again help. For every 2-dimensional
subspace containing v1, the sum of squared lengths of the projections onto the subspace
equals the sum of squared projections onto v1 plus the sum of squared projections along a
vector perpendicular to v1 in the subspace. Thus, instead of looking for the best 2-
dimensional subspace containing v1, look for a unit vector; call it v2, perpendicular to v1
that maximizes |Av| 2 among all such unit vectors. Using the same greedy strategy to find
the best three and higher dimensional subspaces, defines v3, v4, . . . in a similar manner.
This is captured in the following definitions. There is no apriori guarantee that the greedy
algorithm gives the best fit. But, in fact, the greedy algorithm does work and yields the
best-fit subspaces of every dimension as we will show. The second singular vector, v2, is
defined by the best fit line perpendicular to v1 v2 = arg max v⊥v1,|v|=1 |Av| . The value
σ2 (A) = |Av2| is called the second singular value of A. The third singular vector v3 is
defined similarly by v3 = arg max v⊥v1,v2,|v|=1 |Av| and so on. The process stops when
we have found v1, v2, . . . , vr as singular vectors and arg max v⊥v1,v2,...,vr |v|=1 |Av| =
0. If instead of finding v1 that maximized |Av| and then the best fit 2-dimensional
subspace containing v1, we had found the best fit 2- dimensional subspace, we might
have done better. This is not the case. We now give a simple proof that the greedy
algorithm indeed finds the best subspaces of every dimension.
4.4.2. Power Method for Computing SVD
Computing the singular value decomposition is an important branch of numerical
analysis in which there have been many sophisticated developments over a long period of
time. Here we present an “in-principle” method to establish that the approximate SVD of
a matrix A can be computed in polynomial time. The reader is referred to numerical
analysis texts for more details. The method we present, called the Power Method, is
simple and is in fact the conceptual starting point for many algorithms.
4.4.3. Multi-Resolution Singular Value Decomposition
Multi-resolution singular value decomposition is very similar to wavelets
transform, where signal is filtered separately by low pass and high pass finite impulse
response (FIR) filters and the output of each filter is decimated by a factor of two to
achieve first level of decomposition. The decimated low pass filtered output is filtered
separately by low pass and high pass filter followed by decimation by a factor of two
provides second level of decomposition. The successive levels of decomposition can be
achieved by repeating this procedure. The idea behind the MSVD is to replace the FIR
filters with singular value decomposition.
4.4.4. Fusion by MSVD
One can observe that the modification of the present scheme is the use MSVD instead of
wavelets or pyramids. The images to be fused are decomposed into L (l =1, 2... L) level
using MSVD. At each decomposition level (l =1, 2... L), the fusion rule will select the
larger absolute value of the two MSVD detailed coeficients, since the detailed
coefficients correspond to sharper brightness changes in the images such as edges and
object boundaries etc. These coefficients are fluctuating around zero. At the coarest level
(l = L) , the fusion rule take average of the MSVD approximation coefficients since the
approximation coefficents at coarser level are the smoothed and subsampled verion of the
original image. Similalrly, at each decomposition level (l =1, 2... L), the fusion rule takes
the average of the two MSVD eigen matrices
4.5. Applications of SVD
Noise Reduction
Image Compression
Image Forensic Tracks
steganography,
authentication,
labeling,
captioning
fingerprinting
copy control for DVD
hardware/ software watermarking
executable watermarks
signaling (signal information for automatic counting) for propose of
broadcast monitoring count
CHAPTER-5
INTRODUCTION TO MATLAB
sub window in the desktop. M-files are denoted by the extension .m, as in pixelup.m. The
MATLAB editor window has numerous pull-down menus for tasks such as saving,
viewing, and debugging files. Because it performs some simple checks and also uses
colour to differentiate between various elements of code, this text editor is recommended
as the tool of choice for writing and editing M-functions.
To open the editor, type edit at the prompt opens the M-file filename.min an editor
window, ready for editing. As noted earlier, the file must be in the current directory, or in
a directory in the search path.
5.2 Getting Help
The principal way to get help online is to use the MATLAB help browser, opened
as a separate window either by clicking on the question mark symbol (?) on the desktop
toolbar, or by typing help browser at the prompt in the command window. The help
Browser is a web browser integrated into the MATLAB desktop that displays a Hypertext
Mark-up Language (HTML) documents. The Help Browser consists of two panes, the
help navigator pane, used to find information, and the display pane, used to view the
information. Self-explanatory tabs other than navigator pane are used to perform a search.
For example, help on a specific function is obtained by selecting the search tab,
selecting Function Name as the Search Type, and then typing in the function name in the
Search for field. It is good practice to open the Help Browser at the beginning of a
MATLAB session to have helped readily available during code development or other
MATLAB task.
Another way to obtain for a specific function is by typing doc followed by the
function name at the command prompt. For example, typing doc format displays
documentation for the function called format in the display pane of the Help Browser.
This command opens the browser if it is not already open.
M-functions have two types of information that can be displayed by the user. The
first is called the H1 line, which contains the function name and alone line description.
The second is a block of explanation called the Help text block. Typing help at the
prompt followed by a function name displays both the H1 line and the Help text for that
function in there command window. Occasionally, this information can be more up to
date than the documentation of the M-function in question. Typically look for followed
by a keyword displays all the H1 lines that contain that keyword. This function is useful
when looking for a particular topic without knowing the names of applicable functions.
For example, typing look for edge at the prompt displays the H1 lines containing
that keyword. Because the H1 line contains the function name, it then becomes possible
to look at specific functions using the other help methods. Typing look for edge-all at the
prompt displays theH1 line of all functions that contain the word edge in either the H1
line or the Help text block. Words that contain the characters edge also are detected. For
example, the H1 line of a function containing the word poly-edge in the H1 line or Help
text would also be displayed.
5.3 Saving and Retrieving A Work Session
There are several ways to save and load an entire work session or selected
workspace variables in MATLAB. The simplest is as follows.
To save the entire workspace, simply right-click on any blank space in the
workspace Browser window and select Save Workspace As from the menu that appears.
This opens a directory window that allows naming the file and selecting any folder in the
system in which to save it. Then simply click Save. To save a selected variable from the
workspace, select the variable with a left click and then right-click on the highlighted
area. Then select Save Selection As from the menu that appears. This again opens a
window from which a folder can be selected to save the variable.
To select multiple variables, use shift click or control click in the familiar manner,
and then use the procedure just described for a single variable. All files are saved in the
double-precision, binary format with the extension. Mat. These saved files commonly are
referred to as MAT-files.
For example, a session named, says mywork_2003-02-10, and would appear as
the MAT-file mywork_2003_02_10.mat when saved. Similarly, a saved video called
final video will appear when saved as final_video.mat.
To load saved workspaces or variables, left click on the folder icon on the toolbar
of the workspace browser window. This causes a window to open from which a folder
containing MAT-file or selecting open causes the contents of the file to be restored in the
Display the plotting tools from the View menu or by clicking the plotting tools
icon in the figure toolbar, as shown in the following picture
New script
CHAPTER-6
RESULTS
Brain-Axial
(c)Fused image
Brain-Hemisphere
(c)Fused image
The structural similarity (SSIM) index is a method for predicting the perceived
quality of digital television and cinematic pictures, as well as other kinds of digital
images and videos. The first version of the model was developed in the Laboratory for
Image and Video Engineering (LIVE) at The University of Texas at Austin and further
developed jointly with the Laboratory for Computational Vision (LCV) at New York
University.
SSIM is used for measuring the similarity between two images. The SSIM index
is a full reference metric; in other words, the measurement or prediction of image quality
is based on an initial uncompressed or distortion-free image as reference. SSIM is
designed to improve on traditional methods such as peak signal-to-noise ratio (PSNR)
and mean squared error (MSE).
6.1.3. Entropy
H i 1 pi log pi
n
p
i 1
i 1 , 0 ≤ pi ≤ 1
31.591840 0.332138
1. Brain-hemisphere 31.434193 1.876440 0.688603
30.461982 0.670912
2. Brain-axial 30.794717 1.634661 0.521033
CHAPTER-7
CONCLUSION AND FUTURE SCOPE
7.1 CONCLUSION
Image fusion is a technique that combines two or more images relevant
information from a single image and it contains all the information regarding those input
images. The images are taken from multi sensor images of the same scene or multi scene
images from same sensor. The fusion algorithm was implemented by many techniques
like DCT, DWT, and IHS etc. When compared with all techniques the combination of
NSCT and SVD gives effective result. This technique split the images into two or more
up to n. From the comparison on the basis of various performance metrics, it has been
concluded that proposed work performs effectively over existing NSCT and SVD
algorithms.
7.2 FUTURE SCOPE
Image fusion means the combining of multiple images into a single image that has
the most information contented without producing facts which can be missing in certain
image. The design of image fusion in multi-focus cameras to combine data from various
images of the related landscape in order to take the multi focused image.
The proposed work has been implemented and designed in the MATLAB. The
Image fusion using NSCT and SVD provides a future assistance where image is not only
fused by gray image but also it can be applied for RGB image. The results indicate that
the NSCT provides better performance than competing transform such as the curvelet
transform.
REFERENCES:
(1)
(2) A. Cardinali and G. P. Nason, “A statistical multiscale approach to image
segmentation and fusion,” in Proc. Int. Conf. Information Fusion, Philadelphia,
PA, USA, 2005, pp. 475–482.
(3) Ding Li ”Remote sensing Image fusion based on NSCT & PCA”, copy rights
2009.
(4) G. Bhatnagar, Q. M. J. Wu, and B. Raman, “Real time human visual system based
framework for image fusion,” in Proc. Int. Conf. Signal and Image Processing,
Trois-Rivieres, Quebec, Canada, 2010, pp. 71–78.
(5) Kurakula Sravya, Dr. P. Govardhan, Naresh Goud M, “Image Fusion on Multi
Focused Images using NSCT”, International Journal of Computer Science and
Information Technologies, Vol. 5 (4) , 2014.
(6) A. Toet, L. V. Ruyven, and J. Velaton, “Merging thermal and visual images by a
contrast pyramid,” Opt. Eng., vol. 28, no. 7, pp. 789–792, 1989.
(7) V. S. Petrovic and C. S. Xydeas, “Gradient-based multiresolution image fusion,”
IEEE Trans. Image Process., vol. 13, no. 2, pp. 228–237, Feb. 2004.
(8) H. Li, B. S. Manjunath, and S. K. Mitra, “Multisensor image fusion using the
wavelet transform,” Graph Models Image Process., vol. 57, no. 3, pp. 235–245,
1995.
(9) A. Toet, “Hierarchical image fusion,” Mach. Vision Appl., vol. 3, no. 1, pp. 1–11,
1990.
(10) G. Bhatnagar and B. Raman, “A new image fusion technique based on directive
contrast,” Electron. Lett. Comput. Vision Image Anal., vol. 8, no. 2, pp. 18–38,
2009.
(11) Q. Zhang and B. L. Guo, “Multifocus image fusion using the nonsubsampled
contourlet transform,” Signal Process., vol. 89, no. 7, pp. 1334–1346, 2009.
(12) Y.Chai, H. Li, and X. Zhang, “Multifocus image fusion based on features
contrast of multiscale products in nonsubsampled contourlet transform domain,”
Optik, vol. 123, pp. 569–581, 2012.
(13) G. Bhatnagar and Q. M. J.Wu, “An image fusion framework based on human
visual system in framelet domain,” Int. J. Wavelets, Multires., Inf. Process., vol.
10, no. 1, pp. 12500021–30, 2012.
(14) S. Yang, M. Wang, L. Jiao, R. Wu, and Z. Wang, “Image fusion based on a new
contourlet packet,” Inf. Fusion, vol. 11, no. 2, pp. 78–84, 2010.
APPENDIX:
Matlab code:
clc;
clearall;
closeall;
nlevels=1;
dfilter='dmaxflat7';
pfilter='maxflat';
im1=imread('11.png');
figure
imshow(im1);
title('input image 2');
% im1=rgb2gray(im1);
im=imread('12.png');
figure
imshow(im);
title('input image 1');
% im=rgb2gray(im);
im=double(im);
coeffs = nsctdec( double(im), nlevels, dfilter, pfilter );
figure;
imshow(mat2gray(coeffs{1,1}));
title('NSCT coeffs{1,1}');% subplot(1,3,1);
figure;
imshow(mat2gray(coeffs{1,2}{1,1}));
title('NSCT coeffs{1,2}{1,1}');
figure;
imshow(mat2gray(coeffs{1,2}{1,2}));
title('coeffs{1,2}{1,2}');
im1=double(im1);
figure;
imshow(mat2gray(coeffs1{1,1}));
title('NSCT coeffs1{1,1}');
% subplot(1,3,1);
figure;
imshow(mat2gray(coeffs1{1,2}{1,1}));
title('NSCT coeffs1{1,2}{1,1}');
% subplot(1,3,2);
figure;
imshow(mat2gray(coeffs1{1,2}{1,2}));
title('NSCT coeffs1{1,2}{1,2}');
%fusion starts
X.LL = 0.5*(X1.LL+X2.LL);
D = (abs(X1.LH)-abs(X2.LH)) >= 0;
X.LH = D.*X1.LH + (~D).*X2.LH;
D = (abs(X1.HL)-abs(X2.HL)) >= 0;
X.HL = D.*X1.HL + (~D).*X2.HL;
D = (abs(X1.HH)-abs(X2.HH)) >= 0;
X.HH = D.*X1.HH + (~D).*X2.HH;
%apply IMSVD
coe{1,1} = IMSVD(X,U);
%fusion starts
X.LL = 0.5*(X1.LL+X2.LL);
D = (abs(X1.LH)-abs(X2.LH)) >= 0;
X.LH = D.*X1.LH + (~D).*X2.LH;
D = (abs(X1.HL)-abs(X2.HL)) >= 0;
X.HL = D.*X1.HL + (~D).*X2.HL;
D = (abs(X1.HH)-abs(X2.HH)) >= 0;
%apply IMSVD
coe{1,2}{1,1} = IMSVD(X,U);
%fusion starts
X.LL = 0.5*(X1.LL+X2.LL);
D = (abs(X1.LH)-abs(X2.LH)) >= 0;
X.LH = D.*X1.LH + (~D).*X2.LH;
D = (abs(X1.HL)-abs(X2.HL)) >= 0;
X.HL = D.*X1.HL + (~D).*X2.HL;
D = (abs(X1.HH)-abs(X2.HH)) >= 0;
X.HH = D.*X1.HH + (~D).*X2.HH;
coe{1,2}{1,2} = IMSVD(X,U);
%%
imrec=nsctrec(coe,dfilter,pfilter);
x=psnr(imrec,im);
x=x*-1;
fprintf('\nThepsnr is: %f\n',x );
y=psnr(im1,imrec);
y=y*-1;
fprintf('The psnr1 is: %f\n',y );
entrv=entropy(imrec);
fprintf('\n The entropy is: %f\n\n',entrv);
ssimval = ssim(imrec,im);
fprintf('\nThessim value with input image1 is: %f\n',ssimval);
ssimval1 = ssim(imrec,im1);
fprintf('The ssim value with input image2 is: %f\n',ssimval1);
CONTACT DETAILS
Name :
Roll Number. :
Mail Id :
Contact Number :
Name :
Roll Number. :
Mail Id :
Contact Number :
Name :
Roll Number. :
Mail Id :
Contact Number :
Name :
Roll Number. :
Mail Id :
Contact Number :
Name :
Roll Number. :
Mail Id :
Contact Number :