The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008
1226
Figure 1. Synthesized hyperspectral image
keeps the original hyperspectral data’s spectral characteristic
and enhances the degree of image’s transparency and sharpness.
4. CONCLUSIONS
Image fusion for hyperspectral data and high-resolution data is
an emerging technology with the development of remote
sensing imaging technology. It has tremendous application
future on such areas as city planning, environment monitoring,
land utilization, military reconnaissance and so on. However,
most of the current researches and attempts are just tentative.
To convert these theories and techniques into generalized rules
and to apply them to remote sensing area successfully remains a
long way. In this paper, after studying some related algorithms
in the image fusion area, we proposed a wavelet package fusion
algorithm for hyperspectral image based on optimal index band
selection principle and experiments bring us to make the
conclusion that this approach can generate hyperspectral fused
images effectively and robustly.
Figure 2. fusion image Figure 3. part of the fusion image
Information entropy is an effective index to measure the degree
of image’s information; mean gradient can reflect the contrast,
texture characteristic and clearness of images; correlation
coefficients can reflect the similarity between the fused image
and the original hyperspectral image. Thus, we choose
information entropy, mean gradient and correlation coefficients
and other measures to conduct quantitative analysis and the
results in shown in table 1.
Fusion
Algorithm
Information
entropy
Standard
error
Mean
Correlation
coefficient
s
Original
Panchromati
c Image
7.650
57.598
9.529
Original
hyperspectral
image
10.756
29.307
1.777
Fusion image
13.095
57.78
9.566
0.838
ACKNOWLEDGEMENTS
This work described here is partially supported by the grants
from the Open Research Fund of State Key Laboratory of
Information Engineering in Surveying, Mapping and Remote
Sensing, the Key Laboratory of Remote Sensing, Science and
the National Research 863 Program of China(Grant No.
2007AA12Z147).
4.1 References
[1] Zhang Bing, 1997. Hyperspectral data mining supported by
temporal and spatial information, Ph.D. dissertation, Institute of
Remote Sensing Applications of Chinese academy of sciences,
Beijing, pp. 80-88.
[2] Jong-Hyun Park, Ryutaro Tateishi a, Ketut Wikantika,2000,
Multisensor data fusion using multiresolusion analysis. ISPRS,
pp. 987-1002.
[3] Zhang Jun-Ping, Zhang Ye,2002, Hyperspectral image
multi-solution fusion based on local information entropy,
Chinese Journal of Electronics, 11(2), pp. 163-166.
[4] Wilson Terry A, Regers Steven K, Matthew Kabrisky,1988,
Perceptual-based image fusion for hypetspectral data, IEEE
Trans, On G.R.S, 35(4), pp. 1007-1017.
[5] Zhang Jun-Ping, Zhang Ye,2004, Hyperspectral image
classification based on multiple features during multiresolution
fusion, J. Infrared Millim Waves, 23(4), pp. 345-348.
[6] Zhu Shu-Long, Geng Ze-Xun,2000, Wavelet transforms for
remote sensing image processing, Beijing, pp. 24-25.
Table 1. Quantitative analysis results of fused image
From table 1, it can be seen that the information entropy, mean
gradient and correlation coefficients of the fused image by
wavelet package fusion method based on optimal index
principle are greater than the original image. In other words, it
means that the fused image has richer information, greatly