scene for keeping 
If the urban extent 
scene is assigned 
cenes. The scenes 
> included in the 
| with cloud. 
e combination by 
mentally. 
ION 
oints of the subset 
By the 30-km rule, 
2214 COCs. The 
rere determined by 
reject apparently 
contamination less 
failure of SWIR 
tamination, from I 
1e beginning to 31 
e 2214 COCs, the 
cessfully assigned. 
: with condition of 
: search period. In 
of the 263 COCs 
ts of merging the 
to be least-cloud 
7s, the mosaic of 
ganized; however, 
vere incomplete to 
contaminated with 
ted ASTER/VNIR 
manual procedure, 
es were improved; 
    
  
  
  
     
   
   
  
Figure 2. Examples of the results of automated selecting the ASTER/VNIR images. The background images are false color 
composite of the merged ASTER/VNIR. The white lines represents the boundary of the crowd of cities. (a) Tokyo 
(Japan), (b) Sao Paulo (Brazil), (c) Cape Town (South Africa), (d) Najaf (Iraq), (e) Yamoussoukro (Republic of Cote 
d'Ivoire), (f) La Romana (Dominican Republic), (g) Koln (Germany), (h) Rockford (United States of America), (1) 
Gagnoa (Republic of Cote d'Ivoire) 
however, some of them were still contaminated with cloud and 
haze. 
In the selection of satellite images, cloud contamination was the 
main obstacle to determine the best combination of the satellite 
images. In this experiment, we used the database with rate of 
cloud contamination assessed by scene; therefore we could not 
reject cloud contamination partly occurred in the images. By 
assessing cloud contamination for each pixel (e.g. Tonooka et al. 
2010), good-quality pixels of images partly contaminated with 
clouds would be combined with other good-quality pixels. 
Pixel-based assessment also would yield availability of the 
rejected images by less-than-20% criteria to be used for the 
urban area mapping, indicating that the pixel-based assessment 
would be necessary for completing the urban area map of the 
cities of the world. 
4. CONCLUSION 
In this paper, we presented a method to construct coverage 
catalogues of satellite images of urban areas of the world. We 
proposed and implemented an automated algorithm to select the 
best combination of satellite images covering a target COC. The 
experimental results showed that there were considerable spaces 
to be improved for complete coverage of urban area maps, 
especially with assessing cloud contaminations. We found that 
omitted cloud contaminations from the assessment were major 
causes of inferior quality of the merged satellite images. We 
regard that introducing pixel-by-pixel assessment of cloud 
contamination is needed for better quality of the urban area 
maps. 
The method had still much space to be improved; however, we 
regard that, with the improvement, it would be a great 
contribution to completing high-resolution urban maps of the 
world and realizing Global Earth Observation Systems of 
System (GEOSS). 
ACKNOWLEDGMENT 
This research used ASTER Data beta processed by the AIST 
GEO Grid from ASTER Data owned by the Ministry of 
Economy, Trade and Industry of Japan. The study was 
supported by a grant from Grant-in-Aid for JSPS Fellows (22- 
2598).