The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008 
1265 
In this algorithm, threshold value TH needs to be determined. 
The method of choosing the fixed threshold value on experience, 
to be frank, is unreasonable because the threshold value is 
changing along with the system mode, the model deviation, the 
system reference input and the noise and so on, but the fixed 
threshold value is not able to adapt these changes obviously. 
Therefore, we need to enable the system to auto-adapted itself 
to choosing the threshold value. The existing methods of 
choosing the threshold value mainly are: method based on 
statistics, method based on knowledge and method based on 
analytical model [11] 
4.3 Experimental results and appraisal 
In order to confirm the accuracy and the validity of the method, 
we still select the multi-spectrum image and the high spatial 
resolution image of some region terrain landform to carry on 
the fusion experiment. First carry on the strict matching and 
processing, separately carry on the wavelet transformation to it, 
then process according to the method of choosing threshold 
value proposed in this chapter, and the final fusion result image 
is obtained. 
To be compared, these two images have been processed by 
using the traditional wavelet method. The results are as in 
Figure 4. 
( 1 ) Multi-spectral image 
(2) High-resolution image (3) Wavelet image 
(4) large threshold (5) Small threshold (6) Adaptive threshold 
Figure 4 Raw image and fusion image 
(1) is the Landsat TM multi-spectrum image, (2) is the SPOT 
image, (3) is obtained by using the traditional wavelet method, 
(4), (5^ (6) are respectively the fusion images obtained when 
using different threshold value. It can be seen from the figure, 
the fusion images obtained by using the traditional wavelet 
method basically maintain the spectral characteristics of the 
original ones, but the ability to perform spatial details was not 
high; the entire image is quite dim; the edge characteristics like 
river edges are not very clear, the enhancement effect is not 
prominent. Regarding fusion image (4) (threshold value TH =1), 
in fact, this method degenerates the linear method of weighting; 
the spectral characteristic maintenance is insufficient; the 
resolution is not high. But regarding fusion image (5) (threshold 
value TH =0), the ringing effect appears obviously, to cause the 
thing texture fuzzy. But the fusion image (6) obtained in auto- 
adapted threshold value law proposed in the paper is much 
clearer than image (3), (4), (5), either in the region outline or in 
river boundary. Moreover the spectral characteristic also 
maintains well. 
Method types 
spectrum 
tortuosity 
Correlation coefficient 
SPOT 
TM multi- 
Image 
spectral 
Image 
Method Wavelet J=1 
19.4175 
0.7278 
0.7494 
The TH=1 
20.1420 
0.8097 
0.748012 
new TH=0 
24.05 
18.3711 
0.7845 
method Adaptive TH 
0.8551 
0.7969 
Table 2 is parameter contrast of the image fusion result. 
In order to objectively carry on the quality synthetic evaluation 
to this paper method and the traditional wavelet method, the 
two aspect appraisals of the imagery correlation coefficient and 
the spectrum tortuosity have been given. 
Table 2 is parameter contrast of the image fusion result. It can 
be seen from the image parameter result provided in the table 
that the method of this paper all has a very good effect on both 
enhancing the spatial texture characteristic of the multi 
spectrum image and maintaining the spectrum information.