306
12.6
Summary
12: Teerasit Kasetkasem, Pramod K. Varshney
In this chapter, the application of MRF models to image change detection and
image fusion was demonstrated. The ultimate goal for both problems is the
same that is to select the best solution for the problem under consideration
under the statistical framework based on the observed data. Different optimization algorithms were employed depending upon the requirement of each
application.
In image change detection, the simulated annealing algorithm was used to
find the optimum solution. Here, we assumed that the change image and the
noiseless image have MRF properties with different energy functions. Furthermore, we assumed that configurations of changed pixels from different images
given configurations of unchanged pixels are statistically independent. Based
on this model, the MAP equation was developed. The simulated annealing
algorithm was employed to search for the optimum. The results showed that
our algorithm performed fairly well in a noisy environment.
For the image fusion problem, the Gaussian MRF model was employed.
The high-resolution remote sensing image was assumed to have pixel-to-pixel
relation with the fused image, i. e., the observed configurations of any two
pixels in the high resolution image were statistically independent when the
fused image was given. Furthermore, the observed low resolution image was
assumed to be the filtered version of the fused image. The Metropolis algorithm
was employed to search for the optimum solution due to the large number of
possible gray levels (256 in gray-scaled image.) The fused images obtained
from our algorithm dearly showed improvement in the interpretation quality
of the features to be mapped from both multispectral and hyperspectral remote
sensing images.
References
Begas J (1986) On the statistical analysis of dirty pictures. Journal of Royal Statistical Society
B 48(3): 259-302
Bremaud P (1999) Markov chains Gibbs field, Monte Carlo simulation and queues. Springer
Verlag, New York
Bruzzone L, Prieto DF (2000) Automatic analysis of the difference image for unsupervised
change detection. IEEE Transactions on Geoscience and Remote Sensing 38(3): 11711182
Burt PJ, Kolczynski RJ (1993) Enhance image capture through fusion. Proceedings of 4th
International Conference on Computer Vision, 7(4): 593-600
Chair Z, Varshney PK (1986) Optimal data fusion in multiple sensor detection systems. IEEE
Transactions on Aerospace and Electrical Systems AES-22: 98-101
Daniel MM, Willsky AS (1997) A multiresolution methodology for signal-level fusion and
data assimilation with applications to remote sensing. Proceedings IEEE 85(1): 164-180
Geman S, Geman D (1984) Stochastic relaxation, Gibbs distributions and the Bayesian
restoration of images. IEEE Transactions on Pattern Analysis and Machine Intelligence
PAMI-6(6): 721-741
12.6
Summary
12: Teerasit Kasetkasem, Pramod K. Varshney
In this chapter, the application of MRF models to image change detection and
image fusion was demonstrated. The ultimate goal for both problems is the
same that is to select the best solution for the problem under consideration
under the statistical framework based on the observed data. Different optimization algorithms were employed depending upon the requirement of each
application.
In image change detection, the simulated annealing algorithm was used to
find the optimum solution. Here, we assumed that the change image and the
noiseless image have MRF properties with different energy functions. Furthermore, we assumed that configurations of changed pixels from different images
given configurations of unchanged pixels are statistically independent. Based
on this model, the MAP equation was developed. The simulated annealing
algorithm was employed to search for the optimum. The results showed that
our algorithm performed fairly well in a noisy environment.
For the image fusion problem, the Gaussian MRF model was employed.
The high-resolution remote sensing image was assumed to have pixel-to-pixel
relation with the fused image, i. e., the observed configurations of any two
pixels in the high resolution image were statistically independent when the
fused image was given. Furthermore, the observed low resolution image was
assumed to be the filtered version of the fused image. The Metropolis algorithm
was employed to search for the optimum solution due to the large number of
possible gray levels (256 in gray-scaled image.) The fused images obtained
from our algorithm dearly showed improvement in the interpretation quality
of the features to be mapped from both multispectral and hyperspectral remote
sensing images.
References
Begas J (1986) On the statistical analysis of dirty pictures. Journal of Royal Statistical Society
B 48(3): 259-302
Bremaud P (1999) Markov chains Gibbs field, Monte Carlo simulation and queues. Springer
Verlag, New York
Bruzzone L, Prieto DF (2000) Automatic analysis of the difference image for unsupervised
change detection. IEEE Transactions on Geoscience and Remote Sensing 38(3): 11711182
Burt PJ, Kolczynski RJ (1993) Enhance image capture through fusion. Proceedings of 4th
International Conference on Computer Vision, 7(4): 593-600
Chair Z, Varshney PK (1986) Optimal data fusion in multiple sensor detection systems. IEEE
Transactions on Aerospace and Electrical Systems AES-22: 98-101
Daniel MM, Willsky AS (1997) A multiresolution methodology for signal-level fusion and
data assimilation with applications to remote sensing. Proceedings IEEE 85(1): 164-180
Geman S, Geman D (1984) Stochastic relaxation, Gibbs distributions and the Bayesian
restoration of images. IEEE Transactions on Pattern Analysis and Machine Intelligence
PAMI-6(6): 721-741