stanbul 2004 
CED image 
| IRMSS B7 
ge marked A 
. data fusion 
and C show 
details than 
to B7; From 
milar spatial 
ion in image 
  
) B4, IRMSS 
nage B is BT; 
' method 
arly show the 
fusion image. 
ur composite 
mage derived 
. C is a data 
nage A, both 
s some small 
n method can 
prove image 
ge C had the 
different from 
‚ethod can be 
ver resolution 
sthod cannok 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004 
Obviously, PSF method is better than HIS method in keeping 
spectral fidelity. 
  
t 
$ 
Figure 2. Three Band colour composite image of IRMSS 
images and their data fusion image based on CCD image: 
image A is a colour image of B6,B7,B8; image B is a data 
fusion image by HIS method: image C is a data fusion image by 
PSF method. 
[Image spectral profile can also show the effect of data fusion. 
An image spectral profile located at a black line on image B in 
Fig. l, is shown in Fig. 3. Here black curves are the original 
Image spectral profiles: upper curve is B4 image spectral profile, 
the other black curve is B7 image spectral profile; gray curve is 
à spectral profile of data fusion image derived from B7 and B4 
by PSF method. In Fig. 3 it shows clearly that the spectral curve 
of B4 image is far away from the spectral curves of B7 image 
and the data fusion image, while the curve of data fusion image 
IS very close to the curve of B7 image, the difference is only 
that the spectral curve of data fusion image has more high- 
frequent change than the curve of B7 image, and the high- 
frequent change is quite similar to that in the curve of B4 image, 
Which indicates that data fusion image has more spatial details 
and its spectral properties are close to its original image's. 
The image mean and deviation can also show the effect of data 
fusion. The image mean and deviation of original B6,B7,B8 
Mage, data fusion images derived from B6,B7,B8 and B4 by 
PSF method and HIS method respectively are listed in Table 2. 
From table 2, the image deviation of data fusion image is larger 
than the deviation of original image at correspondence band, 
913 
while the deviation of data fusion image produced by PSF is 
similar to the deviation of image by HIS method, which 
indicates that both methods can be used for improving lower 
resolution image; the image mean of data fusion image by PSF 
method is nearly the same as the mean of original IRMSS 
image, but the image mean of the data fusion image by HIS 
method is different from the mean of original IRMSS image, 
which confirms that total spectral energy in data fusion image 
produced by PSF method is the same as that in the original 
IRMSS image and the spectral fidelity of the IRMSS image is 
preserved after data fusion, but the same phenomenon does not 
exist in HIS method. 
Grey Values 
  
  
  
  
  
l0 20 30 40 30 65. 70 20 90... 100. HO 
sample Pixels 
Figure 3. Image spectral profile of samples 
Table 2. The statistics of B6,B7,B8 images and their data fusion 
images 
  
Mean Deviation 
  
B6 B7 B8 B6 B7 B8 
IRMSS | 123 111 75 12.8 107 7.6 
  
  
PSF 123 111 75 259 |] 200 | 157 
  
  
  
  
  
  
  
  
  
HIS * 177 166 130, 35.3.1 29.8.1 24.7 
Note: PSF ': PSF data fusion image; 
HIS ^: HIS data fusion image. 
  
Images, image spectral profiles and statistics indicated that PSF 
method can be used for improving the lower resolution image 
and enhancing the spatial details, and at the same time, the data 
fusion image produced by PSF still preserves the fidelity to the 
lower resolution image spectral properties. Preserving spectral 
fidelity is very useful and important for integrating and 
analysing IRMSS images. 
Since the spatial information in data fusion image is derived 
from the higher resolution image, the image with more spatial 
information content and more spatial details selected as the 
higher resolution image in data fusion processing, will make the 
data fusion image have more spatial details. CCD B5 image has 
more spatial information content and more spatial details 
because of its panchromatic spectral properties, therefore, using 
CCD BS as a higher resolution image will lead to produce a 
good quality data fusion image. At absent of CCD B5, CCD B4 
image is a good image as the higher resolution image to take 
part in data fusion processing. The intensity component of three 
CCD band colour composite image will have more spatial 
details, it will be better to use the intensity component as the 
higher resolution image. Here it is not discussed in detail for 
simplicity.