228 229 
So far simulation activities have been limited to system 
definitions (e.g. SARSIM at ESA). New pieces of software 
(e.g. simulations from Kansas University) indicate that in view 
of simulating SAR scenes, a significant effort has to be 
developed, in terms of instrumentation: 
- at the level of setting up reliable microwave data banks 
(e.g. scattering coefficients versus frequency, 
polarisation, incidence. angle, etc...) 
- at the level of computing time, to produce a simulated scene 
in a practical time. 
it.is clear .thaet.this point is s key.issue for.the near. future. 
3,5 Extraction of Information 
  
Resulting from a coherent process, SAR images feature 
multiplicative noise (speckle). This speckle prevents one from 
using filtering techniques developed for other sources of 
imagery (e.g. Landsat). 
  
Two trends have been identified, in particular during the 
Frascati and Üttawa meetings: 
- the development of general tools (e.g. segmentation 
techniques) for images featuring speckle 
- the development of specific, application oriented filtering 
techniques (e.g. adaptive filtering at CCRS, wave spectrum 
extraction gt MAL and APL etc...) 
These long term developments, based on theoretical studies and 
modelling by (amongst others), V.S. Frost, R.K. Raney and 
Steve Rotheram, are not only aimed at a better understanding of 
the information content of SAR images but also subject of 
practical investigations.(e.g. tracking.of.spectral neaks at 
APL, image spectrum to wave spectrum algorithm facility at 
ESRIN-EPO Frascati, etc...). Up to date information in that 
respect will be reported during the sessions of Rio de Janeiro. 
  
3.6 Radargrammetry and Merging of Multi-Sensor Data 
Radargrammetry deserves particular attention. Although 
| specific equipment for radargrammetric work does not exist, it 
has been shown (Ottawa meeting) that existing analytical 
| 
  
plotters, orthophoto instruments and digital image processing 
systems can be successfully employed for work with radar 
images. Work in this field is also subject of continuous 
updating at every meeting. 
Although they constitute two different areas, we use to relate 
radargrammetry and use of multi-sensor data as the latter 
imposes more constraints on geometrical aspects (e.g. use of 
digital terrain models).