The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008
1173
NIR
-
0.9854
-
-
0.9843
B
0.6396
0.7443
0.8333
0.8530
0.8620
UIQI °
R
0.6870
0.7856
0.8556
0.8586
0.8710
0.7607
0.8186
0.8713
0.8712
0.8910
NIR
-
0.8535
-
-
0.8676
Table3. Quantitative analysis of different fusion methods
5. CONCLUSION
In ideal condition, a good image fusion method tries to
generate the image which a sensor would obtain if it had the
same spectral response of the original MS sensor but the spatial
resolution of the PAN sensor. As a result, to preserve spatial
details and minimize spectral distortion, the spectral
characteristics of sensors have to be taken into account. The
proposed method is based on the generalized fast IHS fusion
framework and two improvements are proposed by considering
sensor spectral response. As shown from the different fusion
experiments, the proposed method has a superior
comprehensive performance and performs better than other IHS
fusion methods regarding both spectral and spatial quality. It is
suitable for various satellite images and extends traditional
three-order transformations to an arbitrary order.
Figure3. Part of Landsat7 ETM+ MS image and fusion result: RGB (743) combination, (a) Original MS image; (b) Fused by IHS
method; (c) Fused by proposed method
Figure4. Part of EO-1 ALI MS image and fusion result: RGB (432) combination, (a) Original MS image; (b) Fused by IHS method;
(c) Fused by proposed method
express sincere gratitude to Doctor J. L. Li for some very
helpful comments.
This work was supported by the Geospatial Information Science
and Technology program (IRT 0438). The author would like to
thank the United States Geological Survey and the Space
Imaging Company for providing images for research and
ACKNOWLEDGMENTS