1134
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Voi. XXXVII. Part Bl. Beijing 2008
matching method adapted to images with various grey aberrations
and random noise. Experiments for matching validity are
implemented by various landscapes images, including farmland,
road, town, river, stream and city. Some images have abundance
features, higher S/N ratio and smaller non-linear variation of pixel
intensity. Some images have bigger variation of pixel intensity and
even grey reversals because of these images were captured in
different season. Also, there are some images with many self
similar areas. Table 4 gives success rate for different scene images
using MI and cross-correlation methods.
Table 4. Comparison of success rates by 16 grey level normalized
mutual information and cross-correlation approaches for different
scene images
Various Scene
images
Cross-correlation
(%)
16 grey levels
normalized
mutual
information (%)
Village 1
61
78
Town
85
93
Farmland
28
69
River
63
81
City
65
88
Village
10
57
Stream
82
100
Average
56.3
80.9
Experiments manifest that the success rate of MI is much greater
than that of cross-correlation method when the scene images have
great grey aberration and even reversal. It shows that the
performance of MI is far excelled than that of cross-correlation in
dissimilar scene matching.
5. CONCLUSION
The performance of MI has no strong relationship to S/N ratio and
information content of images to be matched, but has a strong
relationship to self-similar pattern in the reference image that also
validates the theoretical essence of MI definition and accounts for
MI has strong ability to overcome grey distortion. It is also showed
that good matching performance can be derived even images to be
matched have much lower S/N ratio and grey reversion. Various
scene images are used to test the matching performance based on
MI, and success rates are all higher. It is manifested that MI is a
universal similarity measure and no need feature detection, pre
processing, user initialization and tune of parameter before
matching. It is especially suitable for dissimilar images matching
and it outperforms greatly than cross-correlation method.
REFERENCE
An, R., Jin, X.L., Wang, H.L., Feng, X.Z., Xu, D.X., 2005. Image
ability to obtain correct matching based on feature matching [J].
Infrared and Laser Engineering, 34(4), pp, 469-473.(In Chinese)
Arlene A. Cole-Rhodes, Kisha L. Johnson, Jacqueline LeMoigne,
Ilya Zavorin, 2003. Multiresolution registration of remote sensing
imagery by optimization of mutual information using a stochastic
gradient [J], IEEE Transactions on Image Processing, 12(2),
pp.1495-1511.
Brown, L. G., 1992. A survey of image registration techniques.
ACMComput.Surv, 24(4), pp. 325-376.
Chen, H. M., Varshney, P. K., Arora, M. K., 2003. Performance of
mutual information similarity measure for registration of
multitemporal remote sensing images[J], IEEE Transactions on
Geoscience and Remote Sensing, 41(1 l),pp.2445-2454.
Du, J., Zhang, T.X., 2003. Selection of matching region for scene
matching [J]. Infrared and Laser Engineering, 32(4), pp.368-371.
(In Chinese)
Josien, P.W., Pluim, J. B., Antoine, M., Max, A. V., 2003.Mutual
information based registration of medical images: a survey [J],
IEEE Transactions on Medical Imaging, pp. 1-21.
Knops, Z.F., Maintz, J.B.A., 2006.Normalized mutual information
based registration using k-means clustering and shading correction
[J]. Medical Image Analysis, 10,pp.432~439.
Maes, F., Collignon, A., Vandermeulen, D., Marchal, G., Suetens,
P., 1997. Multimodality image registration by maximization of
mutual information [J]. IEEE Transactions on Medical Imaging,
16, pp.187-198.
Ma, L., Miao, D., Shen, X.B., 2001. A selection criterion of scene
matching field based on image self-matching[J]. AEROSPACE
SHANGHAI, 1,pp.50-54. (In Chinese)
Studholme C., Hill, D. L. G., Hawkes, D. J., 1999. An overlap
invariant entropy measure of 3D medical image alignment[J],
Pattern Recognition, 32(1), pp.71 - 86.
Su K., Jin, S.L., Liu, J.N., 2000. Intelligent scene-matching
positioning technology [J], Infrared and Laser Engineering , 29(1),
pp.1-4.
Viola, P., Wells, W.M., 1997. Alignment by maximization of
mutual information [J], International Journal of Computer Vision,
24, pp. 137-154.
Xie, Y. B., Zhang, W. Q., Jia, Y. L., 1997. Analysis of similarity-
pattern in scene -matching [J], Infrared and Laser Engineering,
26(6), pp.4~9. (In Chinese)
Zitova, B., Flusser, J., 2003. Image registration methods: a survey
[J]. Image and Vision Computing, 21,pp.977-1000.
ACKNOWLEDGEMENT
This research is supported by National Nature Science Foundation
of China (No.40771137); a grant from the State Key Laboratory of
Remote Sensing Science, Jointly Sponsored by the Institute of Re
mote Sensing Applications, Chinese Academy of Sciences and Bei
jing Normal University, and partially supported by 2006103269
NNSFC