120 * 120 Pixel 
Derived edges Zoom (40 e 40 Pixel) 
  
Original 
En oa 143 4 
CA 
a UMOR 
    
1 
i 
2 
à 
  
  
After 
histogram 
matching 
(C) AGF 1993 
  
  
After 
3*3 SNN- 
filtering 
  
  
  
  
  
  
  
  
  
Figure 2: Radiometric operations for reducing striping characteristics by preserving edges, test site Dubai-City 
with given reference material (like maps at various scales) 
or other data sources (like SPOT or Landsat TM). From 
that it can be generally stated that the extraction of topo- 
graphical objects for mapping purposes by visual and 
interactive means is much more effective and reliable than 
by digital classification methods. 
Radiometric operations applied to MOMS-02 images are 
necessary in order to remove or reduce systematic distur- 
bancies like striping effects which have been detected in 
the case of MOMS-02/D2 (Berger and Kaufmann, 1995). 
This is done by means of a histogram matching algorithm 
or by applying the relatively fast 393, edge preserving 
Symmetrical Nearest Neighbour (SNN)- filter (Wang, 
1994). Figure 2 demonstrates the strength of the latter 
method for an actually homogeneously bright area. Fur- 
thermore, desired features like edges can be enhanced for 
a following visual interpreation, e.g. by applying a Laplace- 
filter. 
The visual interpreation and comparison of MOMS-02 test 
data confirmed the theoretical exspectations. Assuming 
typical contrast conditions, road and water networks can 
be extracted nearly completely. Of course, similar prob- 
lems compared to the aerial photo interpretation occur if 
these objects are covered by other objects like trees. On 
the other hand it was found that in an urban area contain- 
ing not so much contrast the derivation of building blocks 
or single houses smaller than 20 m * 20 m is difficult or 
impossible. Vice versa, a pixel size of about 2 m would be 
necessary for such tasks (Konecny, 1995). 
Due to the better spatial resolution MOMS-02 offers a 
sharper demarcation of objects with less jaggies compared 
to SPOT or Landsat TM so that a more precise extraction 
of topographical features is facilitated (figure 3). 
Finally, two features of the MOMS-02 system have been 
found very useful for the purpose of object extraction. 
Firstly, the combination of high-resolution with multispec- 
tral data (e.g. via an intermediate transformation into the 
Intensity-Hue-Saturation-color space) offers the operator a 
colored and enriched source. Of course, combinations with 
other data sources (e.g. Landsat TM, radar) are also pos- 
sible. Secondly, the stereoscopic capability facilitates the 
identification of objects that stand out of their surrounding - 
not only mountains and hills, but also high buildings. 
In summary, MOMS-02 data offer some progress com- 
pared to SPOT or Landsat TM with regard to the interpret- 
ability of objects which are necessary for mapping at 
medium scales. Depending on the specific application the 
capabilities can be sufficient - like in the case of our test 
sites for mapping at 1 : 25 000 - or not, if for example the 
extraction of single houses is needed. 
746 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996