International Journal of Emerging Technology and Advanced Engineering

Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 5, Issue 8, August 2015)

Review of an Edge Preserving Image Fusion

Sandeep Singh1, Navneet Bawa2, Amarjit Kaur3
1
M.Tech. Scholar, 2Associate Professor, Department Of CSE, PTU Regional Centre ACET, Amritsar, India
3
Assistant Professor, Department of CSE , BCET, Gurdaspur, India
Abstract -- The process of merging two or more images into In satellite imaging, two types of images are available.
a single image is called as image fusion. The need of fusing The panchromatic image developed by satellites is
images is to improve the quality of image that helps to retrieve transmitted with the maximum resolution available and the
the more information from the fused image. Image fusion multispectral data are transmitted with coarser resolution.
usually refers to digital image processing,
This will usually be two or four times lower. At the
but optical and analog image processing also are possible.
This article is about image fusion, need of image fusion and receiver station, the panchromatic image is merged with the
techniques that apply to image fusion. The techniques are multispectral data to convey more information.
PCA based image fusion, IHS transform based image fusion, Many methods exist to perform image fusion. The very
Wavelet transform image fusion and High pass filtering basic one is the high pass filtering technique. Later
technique. techniques are based on Discrete Wavelet Transform,
uniform rational filter bank, and Laplacian pyramid.
Keywords-- PCA, HIS ,Wavelet , High Pass Filtering ,color
distortion, transforms 1.1 Need of Image Fusion?
Multisensor data fusion has become a discipline which
I. INTRODUCTION demands more general formal solutions to a number of
In computer revelation, Multisensor Image fusion is the application cases. Several situations in image processing
process of combining relevant information from two or require both high spatial and high spectral material in a
more images into a single image .The resulting image will single image. This is important in remote sensing.
be more explanatory than any of the input images. However, the instruments are not capable of providing such
In remote sensing applications, the increasing availability information either by design or because of observational
of space borne sensors gives a motivation for different constraints. One possible solution for this is data fusion.
image fusion algorithms. Several situations in image Image fusion methods can be generally categorized into
processing need high spatial and high spectral resolution in two sets - spatial domain fusion and transform domain
a single image. Most of the available equipment is not fusion.
capable of providing such data convincingly. Image fusion The fusion procedures such as averaging, Brovey
methods allow the integration of different information method, principal component analysis (PCA)
sources. The fused image can have complementary spatial and IHS based methods fall under spatial domain
and spectral resolution characteristics. However, the approaches. Another important spatial domain fusion
standard image fusion techniques can garble the spectral technique is the high pass filtering based technique. Here
information of the multispectral data while merging. the high frequency details are injected into up sampled
version of MS images[4]. The drawback of spatial domain
approaches is that they produce spatial distortion in the
fused image. Spectral distortion becomes a negative factor
while we go for further processing, such as classification
problem. Spatial distortion can be very well handled by
frequency domain methodologies on image fusion. The
multiresolution analysis has become a very useful tool for
analyzing remote sensing images. The discrete wavelet
transform has become a very useful instrument for fusion.
Some other fusion methods are also there, such as
Laplacian pyramid based, curve let transform based etc.
These methods show a better performance in spatial and
spectral quality of the fused image compared to other
Fig: Image Fusion spatial methods of fusion.

275
 International Journal of Emerging Technology and Advanced Engineering
Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 5, Issue 8, August 2015)
1.2 Image fusion methods Perceived luminance is a extremely difficult aspect of
The images used in image fusion should beforehand color to represent in a digital format, and this has given rise
be registered. Misregistration is a major source of error in to two systems attempting to solve this issue: HSL and
image fusion. Some well-known image fusion methods are: HSV or HSB (V for value or B for brightness). A third
model, HSI (I for intensity), common in computer vision
 PCA based image fusion applications, attempts to balance the advantages and
 IHS transform based image fusion disadvantages of the other two schemes
 Wavelet transform image fusion
 High pass filtering technique 1.2.3 Wavelet Based Image Fusion
1.2.1 Principal Component Analysis (PCA) A wavelet is a wave-like oscillation with amplitude that
begins at zero, increases, and then decreases back to zero. It
PCA is a mathematical tool which transforms a number can classically be visualized as a "brief oscillation" like one
of correlated variables into a number of uncorrelated might see recorded by a seismograph or heart monitor.
variables. The PCA is used comprehensively in image Generally, wavelets are purposefully crafted to have
compression and image classification. The PCA involves a precise properties that make them useful for signal
mathematical process that transforms a number of processing[1]. Wavelets can be combined, using a "reverse,
correlated variables into a number of uncorrelated variables shift, multiply and integrate" technique called convolution,
called principal components. It computes a compact and with portions of a known signal to extract material from the
optimal explanation of the data set. unknown signal. The word wavelet has been used for
The first principal component accounts for as much of decades in digital signal processing and exploration
the variance in the data as possible and each succeeding geophysics. The equivalent French word ondelette meaning
module accounts for as much of the remaining variance as "small wave" was used by Morlet and Grossmann in the
possible. First principal component is taken to be along the early 1980s.Almost all practically valuable discrete wavelet
direction with the maximum variance. The second principal transforms use discrete-time filter banks[5]. These filter
component is constrained to lie in the subspace panels are called the wavelet and scaling coefficients in
perpendicular of the first. Within this Subspace, this wavelets nomenclature. These filter banks may contain
module facts the direction of maximum variance. The third either finite impulse response (FIR) or infinite impulse
principal component is taken in the maximum variance response (IIR) filters. The wavelets forming a continuous
direction in the subspace perpendicular to the first two and wavelet transform (CWT) are subject to the ambiguity
so on. The PCA is also called as Karhunen-Loève principle of Fourier analysis respective sampling theory:
transform or the Hostelling transform. The PCA does not Given a signal with some event in it, one cannot allocate
have a fixed set of basis vectors like FFT, DCT and concurrently an exact time and frequency reaction scale to
wavelet etc. and its basis vectors depend on the data set. that event[2]. The product of the uncertainties of time and
1.2.2 IHS Fusion Method frequency response scale has a lower bound. Thus, in the
scale gram of a continuous wavelet transform of this signal,
The IHS technique is a standard way in image fusion,
such an event marks an entire region in the time-scale
with the major inadequacy that only three bands are
plane, instead of just one point. Also, discrete wavelet
involved. Originally, it was based on the RGB true color
bases may be reflected in the framework of other forms of
space. It offers the advantage that the separate channels
the uncertainty principle.
outline certain color properties, namely intensity (I), hue
(H), and saturation (S). This precise color space is often 1.2.4 High Pass Filter Approach
chosen because the visual cognitive system of human A high-pass filter (HPF) is an electronic filter that passes
beings tends to treat these three components as roughly high-frequency signals but attenuates (reduces the
orthogonal perceptual axes. amplitude of) signals with frequencies lower than the cutoff
These first two values give the two structures the 'H' and frequency. The actual quantity of attenuation for each
'S' in their names. The height corresponds to a third value, frequency differs from filter to filter. A high-pass filter is
the system's representation of the perceived luminance in usually modeled as a linear time-invariant system[6]. It is
relation to the saturation[3]. sometimes called a low-cut filter or bass-cut filter.

276
 International Journal of Emerging Technology and Advanced Engineering
Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 5, Issue 8, August 2015)
High-pass filters have many usages, such as blocking 5. Combine DWT, PCA Transform Multi level fusion
DC from circuitry delicate to non-zero average voltages or where the image undergoes fusion twice using proficient
RF devices. They can also be used in conjunction with a fusion technique deliver better-quality result .output
low-pass filter to make a band-pass filter. For the high pass image contained both high spatial resolution with high
filtering (HPF) fusion, first the ratio between the spatial quality spectral content. This technique is complex in
resolution of the panchromatic and the multispectral image fusion algorithm. Requisite good fusion technique for
is calculated[8]. A high pass convolution filter kernel is enhanced result.
shaped and used to filter the high-resolution input data with
the size of the kernel based on the ratio. The HPF image is III. CONCLUSION
added to each multispectral band. Before the summation, The goal of image fusion is to combine related material
the HPF image is weighted relative to the global standard from two or more source images into one single image such
deviation of the multispectral bands with the weight issues that the single image contains most of the information from
again calculated from the ratio[7]. As a last step, a linear all the source images. In this proposed work previously we
stretch is applied to the new multispectral image to match have used wavelet transformation but because of some
the mean and standard deviation values of the original input problems like shift variant there’s a need to introduce
multispectral image. It displays acceptable results also for algorithm which eliminates such problems. So we have
multi sensorial and multi temporal data. Sometimes the implemented laplacian pyramid with PCA. Laplacian
edges are stressed too much. pyramid smoothens the image and PCA to fuse the image.
II. FUTURE WORK REFERENCES
In computer vision, Multisensor Image fusion is the [1] Shi W, Zhu CQ, Tain Y and J. Nichol, Wavelet-based image fusion
and quality assessment, Int. J. Applied Earth Observation and
process of combining related information from two or more Geoinformation, 6, 241-251 (2004).
images into a single image. The resulting image will be [2] F. W Acerbi-Junior, J. G. P. W. Clevers, ME Schaepman, The
more useful than any of the input images.in research field assessment of multi-sensor image fusion using wavelet transforms
many processes are used for image fusion. It is often not for mapping the Brazilian Savanna, Int. J. Applied Earth Observation
possible to get an image that contains all applicable objects and Geoinformation, 8, 278-288 (2006).
in focus One way to overcome this difficulty is image [3] Miloud Chikr El-Mezouar, Nasreddine Taleb, Kidiyo Kpalma, and
Joseph Ronsin ,“An IHS-Based Fusion for Color Distortion
fusion, in which one can attain a series of pictures with Reduction and Vegetation Enhancement in IKONOS Imagery”,
different focus settings and fuse them to yield an image IEEE Transactions on Geoscience And Remote Sensing, vol. 49, No.
with stretched depth of field. Image fusion techniques can 5, May 2011
increase the excellence and increase the use of these data. [4] Vijay Kumar et al.,“Importance of Statistical Measures in Digital
Image Processing”, ISSN 2250-2459, Volume 2, Issue 8, August
1. Guileless in this is the simplest technique of image 2012.
fusion. The main disadvantage of Pixel level method is [5] Krista Amolins, Yun Zhang, and Peter Dare, “Wavelet based image
that this method does not give guarantee to have a clear fusion techniques—An introduction, review and comparison”,
stuffs from the set of pictures. ISPRS Journal of Photogrammetric and Remote Sensing, Vol. 62,
2. Guileless Extreme in this Resulting in extremely focused pp. 249-263, 2007.
image output obtained from the input image as compared [6] Watkinson, John (1998). The Art of Sound Reproduction. Focal
Press. pp. 268, 479. ISBN 0-240-51512-9. Retrieved March 9, 2010.
to average method. Pixel level method are affected by
[7] Paul M. Mather (2004). Computer processing of remotely sensed
blurring effect which directly effect on the contrast of images: an introduction (3rd ed.). John Wiley and Sons.
the image p. 181. ISBN 978-0-470-84919-4.
3. PCA is a outfit which transforms number of associated [8] Main, Bruce (February 16, 2010). "Cut 'Em Off At The Pass:
variable into number of uncorrelated variables, this Effective Uses Of High-Pass Filtering". Live Sound
property can be used in image fusion.But spatial domain International (Framingham, Massachusetts: ProSoundWeb, EH
Publishing).
fusion my create spectral degradation.
4. The DWT fusion technique may outperform the
slandered fusion method in terms of reducing the
spectral distortion. It offer enhanced signal to noise ratio
than pixel based approach. In this technique final fused
image have a less spatial resolution.

277