The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008
3 73.0405 1682.2747 7.2781 0.7835 0.3869 11.8644
Table 1. The quantitative analysis results of different fusion methods
With respect to image luminance, the mean of the novel method
range between that of the multispectral image and the SAR
image, while the means of the other methods are generally low.
This shows that the novel method well reflects the original
image luminance, which coincides with the subjective analysis
result.
In the aspect of spatial detail information preservation, 1)
Standard deviation: the triangle IHS result has the closest
standard deviation to the original image, and in the next place
the novel method; while the Brovey method exorbitantly
stretched the gray gradient, resulting the image distortion; the
standard deviations of PCA result, Gram-Schmidt result and
normal wavelet transform result are comparatively smaller, as
the distributions of their gray gradient are too concentrated. 2)
Entropy: the entropy of the novel method is larger than that of
any other method, and it is only the novel method result that has
the larger entropy than the original image, which shows that the
novel method has the advantage of preserving information over
other methods. 3) Average gradient: the average gradients of
the PCA result, Gram-Schmidt result and normal wavelet
transform result are comparatively smaller, which shows that
these methods are less able to reflect the small detail
information; Although the average gradient of Brovey result is
the largest, it is larger than that of both the multispectral image
and the SAR image, not truly reflecting the detail information;
The average gradients of the novel method result and triangle
IHS result are between that of the multispectral image and the
SAR image, however, the average gradient of the novel method
result is more closed to that of the SAR image, which means
that the novel method excels triangle IHS in preserving detail
information of SAR image, the very intention we desired.
In the aspect of spectral information preservation, 1)
Correlation coefficient: the correlation coefficients of Brovey
result, triangle IHS result, PCA result with SAR image are
larger than the ones with multispectral image , respectively,
while the situations of normal wavelet transform and the novel
method are reverse. Other than Brovey method, triangle IHS
and PCA are Component Substituting methods (Pradhan,2005),
therefore the correlation coefficients of their result images with
SAR image are naturally larger, so it has no comparability with
the novel method. As for the novel method and normal wavelet
transform, the former has a larger correlation coefficient with
SAR image than the latter, meaning that the novel method is the
better one in preserving information of SAR image. 2) Spectral
distortion: the average spectral distortion of the novel method
result is 65.13, the smallest one, while that of other method
results are over 120, meaning the novel method lost the least
spectral information.
Therefore, synthesized the subject qualitative and objective
quantitative analysis, compare with the traditional methods
concerned in this paper, the novel method not only has obvious
advantage in preserving detail information of SAR image, but
also is better in spectral information preservation and
distribution.
5. CONCLUSIONS
From the application-oriented aspect, this paper proposed auto-
adaptive information preservation fusion method based on
wavelet transform, aiming at better utilizing the information of
SAR image in latter application. It sufficiently considers the
local feature of the image, auto-adaptive fused SAR image and
multispectral image according to the local information capacity,
but not single and ponderous fused regulation for the whole
image.
The experimental result shows that the novel method is better
than the traditional methods adopted in this paper. The novel
method not only better preserves spectral information in
multispectral image, but also better preserves detail information
in SAR image. Especially, the special SAR information is
contained in the novel method fused image. However, the novel
method still has room for improvement.
ACKNOWLEDGMENTS
This work was supported by the Foundation of Fok Ying Tong
Education Foundation (111017), Natural Science Foundation of
China (40671130) and Beijing Natural Science Foundation
(4072016).
REFERENCES
References from Journals:
Deng L., Chen Y., Li J., et al.,2005. Controllable remote
sensing image fusion method based on transform. Journal
Infrared Millimeter and Waves. 24(01), pp. 34-38.
Jia Y.,1998. Fusion of Landsat TM and SAR Images Based on
Principal Component Analysis. Remote Sensing Technology
and Application. 13(01), pp. 46-49.
Li C., Liu L., Wang J., et al.,2004. Comparison of Two
Methods of Fusing Remote Sensing Images with Fidelity of
Spectral Information. Journal of Image and Graphics. 9(11), pp.
1376-1385.
Luclen W., Thierry R. M. M.,1997. Fusion of Satellite Images
of different spatial resolutions: assessing the quality of resulting
images[J]. Photogrammetric Engineering & Remote Sensing.
63(6), pp. 691-699.
Wang R., Qi M., Wang H.,2000. Comparative Study on the
Method of IHS Transformation for Image Fusion. Journal of
The Pla Institute of Surveying and Mapping. 17(04), pp. 269-
272.
Zhang Y., Wu Q.,2006. Information Influence on QuickBird
Images by Brovey Fusion and Wavelet Fusion. Remote Sensing
Technology and Application. 21 (01), pp. 67-70.
References from Books:
1134