2 the 
tran- 
ising 
t the 
> and 
n the 
cen- 
point 
d by 
ge is 
lage 
>, the 
If an 
from 
ler of 
one. 
  
  
  
Fig. 7 (parts of VOYAGER images of Jupiters satellite 
Ganymede) shows the different mosaicking techniques 
(top: copying, middle: weighted elimination of double infor- 
mation, bottom: histogram adjustment and weighted elimi- 
nation) while Fig. 8 demonstrates the capabilities of the 
approach even if the grey levels of the input files vary 
drastically. 
The approach was tested on image mosaics of up to 3000 
images. Fig. 9 shows a mosaic of 779 CLEMENTINE 
images of the Moon containing a few thousands of very 
different types of overlaps. The gaps are due to missing 
input images. 
The overall capabilities of the mosaicking process are 
characterized not only by the quality of the final result but 
also the software performance. Besides the entire auto- 
mation and the robust behaviour on nearly any kind of 
overlapping scenario, the computation time is reduced 
from days or weeks to a few seconds. Even very large 
mosaics with thousands of images can be computed 
within few hours. 
355 
Figure 9: Mosaic of 779 CLEMENTINE images 
  
Figure 8: 11 CLEMENTINE images 
left: images copied 
right: mosaicked with integral histogram adjustment 
and weighted elimination of double information 
  
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B2. Vienna 1996