anbul 2004
> reference
egistration.
wo images
n methods
use of all
scheme of
> MP pair
3 criteria,
rgence has
on to finer
id level is
estimate is
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004
computed by minimizing the SSE between the re-sampled
images gi and image g». This whole process is iterated at each
pyramid level to achieve the final estimation. Gray scale MPs
are created using 3X3 structuring element and then sub-
sampling the filtered image with d — 2. The initial estimated
parameters are identified arbitrarily. Using Levenberg —
Marquardt algorithm by verifying the matching criteria the
parameters ae to a» are iteratively identified in each of the
pyramid levels.
4.4 REGISTRATUION USING GENETICS APPROACH
Randomly initialize the population, sufficiently large to be
representative of the search as a whole. Each chromosome is of
length 32 bits(Prachya, 1999) allocates 12bits for rotation, 10
bits for translation in x-direction and 10 more bits for
translation in y-direction. Each field is a signed magnitude
binary number. A precision factor is used to improve the
accuracy. Evaluate the fitness function for each solution in the
population to see if the termination criteria for optimality are
met. In our case study correlation is used as fitness function.
Used a weighted roulette wheel sampling to reproduce strings
of the next generation in proportion to their fitness. Evaluate
the fitness of each new individual. Thus we obtain a solution
string or chromosome, which is used to transform the image
using affine transformation and bilinear interpolation.
Population size and number of generations were limited to 150,
regstration accuracy observed as less than a pixel.
S. CONCLUSIONS
Wavelet Modulus Maxima appraoch assumes the images are of
same resolution. In this method threshold parameters need to be
interactively provided. Due to pyramidal approach it allows for
faster implementation and higher registering precision. It is
more adequate to register images taken from the same sensor.
It worked well for images taken at different times, which are
typical to remote sensing applications. Since this uses the
control points approach it can rectify the local errors, which
emulates manual registration of images. FFT technique
provides accuracy acceptably good. The algorithm works for
images in which the scale change is less than 1.8 (Hongjie Xie,
2003) Due to the global transform this approach cannot
determine local geometric distortions. The MPIR algorithm
with an intensity-based differential matching technique is
reliable and efficient. This algorithm capable of measuring the
errors, to sub pixel accuracy, the displacement between images
subjected to affine transformation, which includes simultaneous
translation, rotation, scaling, and shearing. GAs can efficiently
search the solution space and gives the solution to achieve the
sub pixel accuracy without identifying the control points.
Through global transformation a model can be established for
translation and rotation errors. The proposed algorithm expects
both the images are of same scale. Computational efficiency
can be improved by adopting the pyramidal approach.
Depending on the type of variations in the medical images of
Computerized Tomography, PET or MRI images some of these
techniques can be adopted for making the various observations.
It is unlikely that a single registration scheme will work
satisfactorily. To characterize these algorithms the common
data sets from IRS PAN are used and there is no scale variation.
ACKNOWLEDGEMENTS
The authors would like to thank Dr. R. R. Naval gund, Director
NRSA for encouraging us to carry out this work at NRSA.
703
REFFERENCES
Alliney, S., Anij Morandi, C. 1986, Digital image registration
using projections IEEE Trans. PAMI-8, 2 (Mar.), 222-233.
A. Morales, T. Acharya, and S. Ko, “Morphological pyramids
with alternating sequential filters”, TEEE Trans. Image
Processing, vol.4, pp. 965-977, 1995.
A.S.Kumar, A.S.Manjunath and K.M.M. Rao , * Merging IRS-
Multispectral and PAN images by A-Trous wavelets". Int.
Journal of Remote Sensing , 2003 (communicated).
B.S.Reddy and B.N.Chatterji, *An FFT-based technique for
translation, rotation and scale invariant image registration",
IEEE Transactions on Image Processing, 5(8): 1266 —
1271,1996.
De Castro E, and Morandi C, 1987. Registration of translated
and rotated images using finite Fourier Transforms, IEEE
Trans. PAMI-9, 5 (Sept.),700-703.
D. E. Goldburg, Genetic Algorithms in Search: optimization
and machine learning, Reading, Mass. Addison- Wesley,1989.
H. Li, B. S. Manjunath, and S. K. Mitra, ^A contour-based
approach to multisensor image registration," IEEE Trans.Image
Processing, vol.4, pp. 320-334, 1995.
Hongjie Xie,NigelHicks,G.randy Keller Haitao Huang ,"An
IDL/ENVI implementation of the FFT-based algorithm for
automatic image registration” .Int.J]] of Computers and
Geosciences, vol .29 ,pp 1045-1055, 2003.
J. H. Holland, *Adaptation in Natural and Artificial System,"
University of Michigan Press, Ann Arbor, 1975.
L.G. Brown, “A survey of image registration techniques,”
ACM Computer Surveys, vol. 24, pp325-376, 1992.
Leila M.Fonesca and Max H.M.Costa,”Automatic Regestration
of Satellite Images”,Proceedings on IEEE transaction of
computer society,1997,pp 219-226.
Prachya Chalermwat, Tarek A. El-Chazawi “Multi resolution
image registration using Genetics, ICIP(2) 1999:452-456.
P. Thevenaz, U.E. Ruttimann, and M. Unser, “A pyramid
approach to sub pixel registration based on intensity,” IEEE
Trans. Image Processing, vol.7, pp. 27-41, 1998.
Q.Zheng and R.A.Chellappa. “Computational vision approach
to image registraion”. IEEE Transactions on Image Processing,
2(3) : 311-326, July 1993.
S.G. Mallat “A theory for multi-resolution signal
decomposition: the wavelet representation.” IEEE Trans. On
Pattern Anal. And Machine Intell. , 11(7):674-693, July 1989.
Y.Keller, A.Averbuch, “FFT Based Image Registration”,IEEE
international conference — ICASSP 2002. Orlando.
YOUNG.D “Straight lines and circles in the log-polar image”.
Proceedings of the British Machine Vision Conference 2000,
BMVC, 2000, Bristol, UK, 11-14 September 2000, pp 426-435
Zhongxiu Hu and Scott T.Acton,"Morphological Pyramid
Image Registration" 4" [EEE south west symposium 2000.
P227.
