International Journal of Scientific & Engineering Research, Volume 5, Issue 12, December-2014 44

ISSN 2229-5518

Different Image Registration Methods –

An Overview
Ms. Ritu Singh Phogat, Mr. Hardik Dhamecha, Dr. Manoj Pandya , Mr. Bharat Chaudhary , Dr. Madhukar Potdar

Abstract The main aim of this paper is to provide a review of classic image registration methods. Registration is a fundamental stage in 3D and 2D
reconstruction process, which is used to match two or more images that can be taken in different moments, from different sensors, or different
viewpoints. Two images are aligned into a common coordinate system in order to monitor the subtle changes between them. Registration algorithms
compute transformations to set correspondence between the two images. Previous surveys of image registration divide the field algorithmically,
focussing on how registration is accomplished. This approach provides a good basis for classifying different techniques, however it does little to
illuminate the conditions important to the problem of registration. Understanding the image registration problem space and the conditions under which
specific algorithms perform best allows for further automation of image registration, and makes the field more accessible to non-experts.

Index Terms— Automatic Image Registration, Feature Detection, Matching, Mapping Function, Resampling, Transformation

——————————  ——————————

image, this is known as matching problem, and is also the

I. INTRODUCTION most time consuming stage during the execution of
algorithm. Second, there is need of transformation in the

I n image processing, for example, when combining the

three dimensional information of one of the images
regarding its coordinate system and its relationship to the

IJSER
information content of image, we are interested in the image that was chosen as its reference. In simple words,
relationship between two or more images. The analysis of image registration process is an automatic or manual
this relationship usually be easily controlled or dealt with procedure which tries to find corresponding points
once a correspondence is set up between the images. The between two images and align them spatially to minimize a
task of setting up this correspondence is called image desired error, that is, a consistent distance measure
registration. Image registration is basically of two types: between two images. There is a rapid development in
1. Image to image registration: Here two or more images image acquisition devices during last few decades and also
are aligned to integrate or fuse corresponding pixels that growing amount and diversity of images that are obtained
represent the same objects. has invoked the research on automatic image registration.
2. Image to map registration: Here input image is warped The intention of our article is to cover relevant approaches
to match the map information of a base image, while introduced and in this way map the current development
retaining the original spatial resolution. Image registration of registration techniques. Methods published before 1992
is the process of calculating spatial transforms which align that became classic or that introduced key ideas, which are
a set of images to a common observational frame of still in use, are included as well to retain the continuity and
reference, often one of the images in the set. Registration is to give complete view of image registration research. We
a key step in any image analysis or understanding task want to summarize main approaches and point out
where different sources of data must be combined. It is a interesting parts of the registration methods instead of
critical component of computational photography [1] going into the details of particular algorithms or describe
remote sensing [2], [3], automatic manufacturing processes, some results of comparative experiments.
and medical image processing [4], [5]. More recently it has
been used to create navigable models of a scene from a II. CLASSIFICATION OF IMAGE REGISTRATION
database of photographs [6], to remove unwanted objects METHODS.
from overlapping images, etc. During registration process Past image registration surveys have provided a
two situations become evident. First, it is impossible to methodological taxonomy for understanding the different
algorithms used to solve various registration problem. The
————————————————
work of Barbara Zitova and Jan Flusser [7] divides the field
• Ritu Singh Phogat, Masters Student in Electronics And Communication,
into two groups which are area and feature based methods.
MEFGI College, Rajkot. Area based methods are preferably applied when in images
• Hardik Dhamecha, Asst.Professor in EC Department, MEFGI prominent details are absent and distinctive information is
College,Rajkot provided by grey levels / colours rather by local shapes
• Bharat Chaudhary and Manoj Pandya,(project scientist), Dr. Madhukar
Potdar, project diretor in BISAG, Gandhinagar. and structure. Feature based matching methods are applied
when there is more local structural information carried by
determine which of the coordinate system points of one image intensities. These methods make use of image
image matches with the corresponding points of another features that are derived from feature extraction algorithm.
IJSER © 2014
http://www.ijser.org
International Journal of Scientific & Engineering Research, Volume 5, Issue 12, December-2014 45
ISSN 2229-5518
Point of sharp variations such as edges, corners, contours, pattern recognition in which the location and orientation of
surfaces, point of intersection etc, what carries valuable a template or pattern is found in picture. Cross correlation
information about images are used for matching. Their is a similarity measure or match metric. For template T and
model includes four steps of image registration, which are: image I, where T is small compared to I, the two
Feature detection, Feature matching, Mapping function dimensional normalized cross-correlation function
design, and Image transformation and resampling. They measures the similarity for each translation. If template
provide a model of variation according to the manner in matches the image, then the peak of cross correlation will
which the images were acquired which include different be at C (i, j). Since local image intensity would influence the
viewpoints, different times / conditions, different sensors, measure so the cross correlation should be normalized.
and finally scene to model registration. Within their survey
they do not use this mapping directly; however in many Observations:
cases they discuss the type of problem each method has The flatness of the similarity measure maxima (due to the
been designed to solve, allowing a similar mapping of self-similarity of the images) and high computational
methodology from situation. The four basic steps of image complexity are the major drawbacks of correlation
registration are as follows: methods. By pre-processing or by using the edge or vector
• Feature detection: The Salient and distinctive correlation the maximum can be sharpened effectively.
objects (Closed boundary regions, Edges,
Contours, Line intersections, Corners, etc) in both B) Feature Based Methods or Point Mapping Method
reference image and sensed image are detected Feature based matching techniques do not use the grey
automatically or manually by a domain expert. values to describe matching entities. It makes use of image
features derived by feature extraction algorithm/process.
• Feature matching: The correspondence between Abstracting large amount of information from original data
the features in the reference and sensed image is input and filter out the redundant information is the main

IJSER
established. Matching method is based on image purpose of feature extraction i.e the information which is
content or on the symbolic description of control no longer needed is filtered out. Features that are uniquely
point-set. found in both images and more tolerant of local distortions
are selected. Computing of proper transformation depends
• Transform model estimation: The parameters on these features. Therefore sufficient number of features
and type of the so-called mapping functions, must be detected to perform calculation. After detecting
aligning the sensed image with the reference features in each image, it is necessary to match them. This
image, are estimated. method is primary approach to register two images whose
type of misalignment is not known. This occurs if class of
• Image resampling and transformation: transformations cannot be easily categorized as translations
By means of the mapping functions the sensed or rigid-body movements. In this we can use landmarks
image is transformed. and match them by using general transformation. The
methods of point mapping consist of three stages which are
as follows:
• Computing features in the images.
•Control points in reference image are corresponded with
feature points in data image.
• Spatial mapping.
Control points for point matching play an important role in
this method. Control points may be corners, line of
intersections, points of locally maximum curvature on
contour lines, centre of windows having locally maximum
curvature and centre of gravity of closed-boundary regions.

Observations:
Feature based registration methods overcomes the
limitation of correlation method and are often in use,
particularly because of their easy hardware
implementation, which makes them useful for real-time
FIG 1: STEPS IN IMAGE REGISTRATION PROCESS applications. Disadvantage of this method refers to the
remarkableness of the window content. There are high
A) Pixel Based Method: chances that a window containing a smooth area without
In this method cross-correlation statistical approach is used any crucial details will be matched incorrectly with other
for registration. It is often used for template matching or smooth areas in the reference image due to its non-saliency.
IJSER © 2014
http://www.ijser.org
International Journal of Scientific & Engineering Research, Volume 5, Issue 12, December-2014 46
ISSN 2229-5518
The features for registration should be preferably detected method has its own limitations. When images are of low
in distinctive parts of the image. Windows may not have resolution, when images contain little information, or when
this property, whose selection is often not based on their the region of overlap is small then mutual information
content evaluation. result in mis-registration .It has one more limitation of
speed , when time is an important constraint one cannot
C) Registration Based on High Level Features use this method . Although it has some limitations entropy
(Contour Based Image Registration) and mutual information are best approaches for
For matching image feature point’s high statistical features multimodal image registration.
are used in this method. For extracting regions of interest
from the image, colour image segmentation is used. For E) Image Registration in Frequency Domain
example, for a given set of colour mean is calculated which Correlation theorem states that, “the Fourier transform of
is denoted as “m”. For segmentation process each RGB the correlation of two images is the product of Fourier
pixel in an image is classified as having a colour in transform of one image and complex conjugate of Fourier
specified range or not. Also the Euclidean distance is transform of other”. The Fourier transform of an image
required to measure similarity. Let ‘z’ be an arbitrary point f(x,y) is a complex function, each function value has real
in RGB space and ‘T’ be the threshold. Euclidean distance part R(ωx,ωy) and an imaginary part I(ωx,ωy)at each
between ‘z’ and ‘m’ is given by as shown below: frequency (ωx,ωy) of frequency spectrum. Shift theorem
guarantees that phase of cross power spectrum is
D (z, m) =║z-m║= [(z-m)T(z-m)]1/2 equivalent to the phase difference between the images. If
we represent the phase of cross power spectrum in its
The locus of points is the sphere of radius T and point that spatial form, i.e. by taking the inverse Fourier transform of
is contained within, or on the surface of the sphere satisfy the representation in the frequency domain, then we will
specified colour criterion. Coding these two sets of points in have a function which is an impulse, that is approximately

IJSER
the image with black and white produces a binary, zero everywhere except at displacement which is needed to
segmented image. Noise is removed after segmentation optimally register two images. Above method is used to
process using Gaussian filter. Thresholds blurred the image register images having only translation.
and then obtain the contour of an image.

Observations:
Feature based methods do not use the gray values for
matching and hence overcomes the limitations of spatial Observations:
methods. Feature based method filter out the redundant It is observed that in frequency based method accuracy is
information. Accuracy of this method is more but the more than correlation method but less as compared to other
limitation is that it is manual and slow. methods. But if we extract image features and then apply
Fourier method accuracy increases. In frequency domain it
D) Multimodal Image Registration Using Mutual should be noted that some form of interpolation must be
Information used.
Registration of multispectral / multisensory images is a
challenging area. In general such images have different F) Image Registration using Genetic Algorithm
gray level characteristics and simple techniques such as John Holland and his colleagues formally introduced
those based on area correlation cannot be applied directly. genetic algorithms (GAs) in paper [12] as mentioned in
This section is an attempt to solve this difficult problem by reference. GA’s are based on the natural concept of
employing a basic concept from information theory. evolution, suggesting that diversity helps to ensure a
According to Manjusha Deshmukh and Udhav Bhosle [11], population survival under changing environmental
it is found that entropy of image does not change even if conditions. Genetic algorithm (GA) is used to match the
histogram has changed. Even after randomly shuffling corresponding features and fit the satellite image on the
pixels of image, entropy of image remains same. Also they vector map by optimizing the transformation accuracy on
concluded that natural images contain less uncertainty. In checked and control points. GA’s have intrinsic parallelism
this type of images pixel intensity values depend on and are iterative procedures. They maintain a population of
neighbouring pixels. In other words, in natural image the P candidate solutions encoded in the form of chromosome
value of pixel is likely to be very close to some of its string of some alphabets, typically the binary symbols "0"
neighbours. Hence this dependency reduces the total and "1". The initial population can be selected heuristically
entropy [10]. or randomly. A fitness function, representing the quality of
each candidate solution, is evaluated for each generation.
Observations: Fit candidates, those with fitness value higher than certain
It is observed that proposed method yields a more accurate threshold, will be selected for the reproduction in the next
registration than any other registration method. But this generation. The selected candidates are combined using a
IJSER © 2014
http://www.ijser.org
International Journal of Scientific & Engineering Research, Volume 5, Issue 12, December-2014 47
ISSN 2229-5518
genetic reproduction operation called “crossover”. The • Employing a Genetic algorithm for conjugate
crossover operator exchanges portions of the chromosome feature identification and modelling.
to produce better candidates with higher fitness in the next
generation. It is necessary to perturb the string of the G) Image Registration using Embedded Maps
chromosome to guarantee that the chances of searching a Xiaoxing Li , Xiaojing Long, Paul Laurienti, Christopher
particular subspace of the problem space is never zero and Wyatt has presented registration via embedded maps
for this task mutation operator is then applied. This (REM), a deformable registration algorithm for images with
prevents the algorithm from becoming trapped in local varying topology. The algorithm represents 3-D images as
optima [13], [14]. The whole population is evaluated again 4-D manifolds in a Riemannian space (referred to as
in the next generation. The process continues until a embedded maps). Registration is performed as a surface
termination criterion is reached. The termination criteria evolution matching one embedded map to another using a
may be finding an acceptable approximate solution, diffusion process. The approach differs from those existing
reaching a specific number of generations, or convergence in that it takes an a priori estimation of image regions
of the solution. where topological changes are present, for example lesions,
and generates a dense vector field representing both the
shape and intensity changes necessary to match the images.
The algorithm outputs both a diffeomorphic deformation
field and an intensity displacement which corrects the
intensity difference caused by topological changes.
Multiple sets of experiments are conducted on magnetic
resonance imaging (MRI) with lesions from OASIS and
ADNI datasets. These images are registered to either a
brain template or images of healthy individuals. An

IJSER
exemplar case registering a template to an MRI with
tumour is also given. The resulting deformation fields were
compared with those obtained using diffeomorphic
demons, where topological changes are not modelled.
These sets of experiments demonstrate the efficacy of our
proposed REM method for registration of brain MRI with
severe topological differences.

III. OTHER RELEVANT SURVEY

Maintz [8] provides insight into the use of registration in
medical imaging, providing important methods and
variations relevant to that field. The taxonomy divides both
algorithmically and based on the modality of the data,
again providing a similar mapping. Pluim et al. [9] also
survey medical imaging, focusing on Mutual-Information-
Based registration techniques. Their taxonomy classifies
algorithms into two main categories: methodological
aspects and aspects of application. E. Parmehr, C.Fraser, C.
Zhang, J. Leach[19] has presented a new hybrid intensity-
based approach that utilizes both statistical and functional
relationships between images, particularly in the case of
registering aerial images and 3D point clouds. An
important advantage of LiDAR systems is direct acquisition
FIG 2: CONCEPT OF GENETIC ALGORITHM of 3D coordinates of ground objects. However, LiDAR data
can have difficulty revealing information along break lines.
Observations:
According to F. Samadzadegan and M. Hoseini [15]
IV. RESULT AND CONCLUSION
automatic registration using genetic algorithm has proved
In this paper we have categorized different
to be very efficient and reliable for satellite imageries based
methods/techniques for image registration, their
on digital vector maps. The implemented methodology has
advantages and also how they are different from other
the following characteristics:
methods has been included. When we register the images
• Utilization of a multi-resolution representation
with local geometric distortions, nonlinear we come across
of information and mathematical models.
two basic problems. One of them is how to match control
points while the second problem is how to choose mapping
IJSER © 2014
http://www.ijser.org
International Journal of Scientific & Engineering Research, Volume 5, Issue 12, December-2014 48
ISSN 2229-5518
functions for registration as sometimes it is difficult to
discriminate the image deformation and real changes
occurred in the scene. An automatic matching method REFERENCES
cannot be used for matching of control points (first [1] A. Agarwala, M. Dontcheva, M. Agrawala, S. Drucker, A. Colburn,
problem) as the deformation between images is arbitrary B. Curless, D. Salesin, and M. Cohen, “Interactive digital
(Zitova and Flusser, 2003). In the future, we need to photomontage,” in SIGGRAPH ’04: ACM SIGGRAPH 2004 Papers,
develop an expert image registration method from a New York, NY, USA, 2004, pp. 294–302.
combination of various approaches which takes care of the
[2] T. M. Lillesand and R. W. Kiefer, Remote Sensing and Image
type of the given task and provides an appropriate Interpretation, 6th ed. Wiley, 2007
solution. In pixel based method cross correlation is used as
similarity measure. It is observed that in natural images [3] J. B. Campbell, Introduction to remote sensing, 4th ed. Guildford
like buildings or scenery, correlation method shows match Press, 2008.
at multiple points. The feature based method makes use of
[4] J. Maintz and M. Viergever, “A survey of medical image
features like point of intersection, edges, corners, centers of
registration”, Medical Image Analysis, vol. 2, no. 1, pp. 1–36, 1998
contours etc. for matching sample template with reference
image. But this method is manual and hence time [5] J. Pluim, J. Maintz, and M. Viergever, “Mutual-informationbased
consuming. The method combining image features with registration of medical images: a survey”, Medical Imaging, IEEE
correlation method have many advantageous properties of Transactions on, vol. 22, no. 8, pp. 986–1004, Aug. 2003. 313
both feature-based and intensity based. It overcomes the
limitation of intensity based method. Contour based [6] N. Snavely, R. Garg, S. M. Seitz, and R. Szeliski, “Finding paths
methods do not use the gray values for matching and hence through the world’s photos,” ACM Transactions on Graphics
(Proceedings of SIGGRAPH 2008), vol. 27, no. 3, pp. 11–21, 2008
overcomes the limitations of spatial methods. Feature based
method filter out the redundant information. Accuracy of [7] B. Zitova and J. Flusser, “Image registration methods: a survey,”

IJSER
this method is more but the limitation is that it is manual Image and Vision Computing, vol. 21, pp. 977–1000, 2003.
and slow. In frequency based method accuracy is more
than correlation method but less as compared to other [8] J. Maintz and M. Viergever, “A survey of medical image
registration,” Medical Image Analysis, vol. 2, no. 1, pp. 1–36, 1998
methods. But if we extract image features and then apply
Fourier method accuracy increases. In frequency domain it [9] J. Pluim, J. Maintz, and M. Viergever, “Mutual-informationbased
should be noted that some form of interpolation must be registration of medical images: a survey,” Medical Imaging, IEEE
used. These are some of the conclusions about methods Transactions on, vol. 22, no. 8, pp. 986–1004, Aug. 2003.
used for registration of images which are in same spectral
[10] Mert Rory Sabuncu, “Entropy-Based image registration, Phd
band. Image registration is difficult when images are
thesis.
obtained through different sensor types. It is observed that
method based on information theory yields a more accurate [11]Manjusha Deshmukh and Udhav Bhosle,”A Survey Of Image
registration than any other registration method. But this Registration”, International Journal of Image Processing (IJIP), Volume
method has its own limitations. When images are of low (5) : Issue (3) , 2011
resolution, when images contain little information, or when
[12] Holland, J., 1975. Adaptation of Natural and Artificial Systems,
the region of overlap is small then mutual information The University of Michigan Press, Ann Arbor.
result in mis-registration. It has one more limitation of
speed, when time is an important constraint one cannot use [13] Goldberg, D.E., 1989. Genetic Algorithms in Search, Optimization
this method. Although it has some limitations entropy and & Machine Learning, Addison-Wesley Longman
mutual information are best approaches for multimodal
[14] Husband, P., 1990. Genetic Algorithms in Optimisation and
image registration. Polynomial transformation was
Adaptation, Advances in Parallel Algorithms Kronsjo and
mentioned as a major tool for transforming all tie points Shumsheruddin ed., pp. 227-276, 1990
from image space to object space. The proposed algorithm
was extensively tested on different brain MR image [15] Farhad Samadzadegan, Sara Saeedi, Mohammad Hoseini,”
datasets carrying minor to severe pathology. The Automatic Image to Map Registration Based on Genetic Algorithm”,
Proceedings of the 2nd WSEAS International Conference on Remote
registration results are compared with those obtained from
Sensing, Tenerife, Canary Islands, Spain, December 16-18, 2006
diffeomorphic demons, which demonstrate the
effectiveness of our proposed algorithm in terms of [16] Xiaoxing Li , Xiaojing Long, Paul Laurienti, Christopher Wyatt,
converging to correct registration in the presence of lesion. and ALZHEIMER’S DISEASE NEUROIMAGING INITIATIVE,”
Registration of Images With Varying Topology Using Embedded
Maps”, IEEE TRANSACTIONS ON MEDICAL IMAGING, VOL. 31,
ACKNOWLEDGEMENT NO. 3, MARCH 2012
The authors wish to thank T.P. Singh, Director,
Bhaskaracharya Institute of Space Application and Geo- [17] S. Kullback, Information theory and statistics. Dover Publications,
Informatics for his valuable support and inputs in the Inc., 1968.
research work.
IJSER © 2014
http://www.ijser.org
International Journal of Scientific & Engineering Research, Volume 5, Issue 12, December-2014 49
ISSN 2229-5518
[18] T.M. Cover and J.A. Thomas, Elements of information theory, John Using a Hybrid Intensity-Based Method”, 978-1-4673-2181-
Wiley & Sons, Inc., 1991 5/12/$31.00 ©2012 IEEE

[19] Ebadat G. Parmehr, Clive S. Fraser, Chunsun Zhang, Joseph

Leach, “Automatic Registration of Aerial Images with 3D LiDAR Data

IJSER

IJSER © 2014
http://www.ijser.org