Mount 
| points 
results 
et 
derived 
el of the 
ed from 
ions do 
can be 
compete 
EMs for 
code the 
ented in 
id a half 
s means 
50° but 
g greater 
and full 
g the 35 
the SAR 
way in 
the full 
opposite 
is fully 
igh high 
ch more 
"e. If 
lark has 
ufficient 
Alternatives 
Although there is considerable interest in interferometric SAR, 
and there is a much higher probability of obtaining good data 
than from stereo, there have as yet been no quantitative results 
published. The same applies to shape from shading. The 
accuracy which may be obtained is therefore unknown. Other 
sources of data such as maps are available but the task of 
digitising contours to form DEMs on a wide scale is daunting 
and the collection of data from different countries and different 
mapping organisations is fraught with administrative and 
political problems . 
Implementation 
The problems of collecting DEM data are clearly large and the 
sensible approach is to make provision to use data from all 
possible sources. This implies the use of an integrated 
approach in which the user is provided with a database, a tool 
box with which to operate on the data and a user interface such 
as a workstation. Such a workstation is being developed at 
UCL, Dowman and Upton (1991). 
The main processes required prior to display (although the data 
may have already been operated on during previous 
processing) are as follows: 
SAR geocoding; 
Merging two data sets; 
Overlaying raster data with vector data. 
The interactive processes required are: 
Read image co-ordinates (mono or stereo); 
Image enhancement; 
Ground control point identification; 
Feature extraction. 
After some of these operations the data may be sent back to be 
added to the data base or to replace existing data. Other data 
will then be subject to further off line processing: 
Calculate orientation or orbit data; 
Calculate object space co-ordinates (stereo 
mode); 
Image matching. 
The display is a crucial part of the workstation. It is the point 
at which the operator interacts with the data and it is essential 
that this is done quickly and efficiently. The display may 
show a single image or it may show two images. If two 
images are required then parallax free stereo viewing must be 
available; this means moving both images independently, 
under control of their respective camera model, and providing 
a stereo viewing system. The following operations must be 
possible: 
Overview whole images; 
Measure ground control points; 
Revise and compile map data. 
Such a system allows great flexibility in processing the data 
and merging it with other data. 
5. CONCLUSIONS 
The full exploitation of SAR data through geocoding is still at 
an early stage. DEMS are crucial to a realisation of the full 
potential of SAR and if these DEMS can be produced from the 
SAR data itself then the use of SAR can become extremely 
effective. It has been shown that a number of techniques are 
available and that each has a potential, it is therefore important 
that development and testing continue and that all techniques 
become available in a user friendly environment. 
427 
REFERENCES 
Day and Muller, 1989. Digital elevation model production 
bystereo matching SPOT stereo pairs: a comparison of 
algorithms. [Image and Vision Computing 7(2):95-102. 
Berry P A M, Birkett C M, Devayya, Gilbert L K, Jackson 
D, Palmer D E, Rapley C G, Wingham D J, Wolf M, 1992. 
The construction and verification of global digital elevation 
models using satellite altimeter observations. Satellite 
Symposium No2, ISY Conference Munich, March 1992. 
Clark C, 1991. Geocoding and Stereoscopy of Synthetic 
Aperture Radar Images. PhD Thesis, University of London. 
Denos M, 1991. An automated approach to stereo matching 
Seasat imagery. Proceedings of British Machine Vision 
Conference, Glasgow, 1991. Springer-Verlag 
Dowman I J and Upton M, 1991. A Sun environment for 
geocoding.  Geocoded products: Intercomparison and 
applications. Proceedings of Third International Workshop on 
Image Rectificaion and Spacebourne SAR. Farnham Castle, 
October 1991. EOS Ltd Fleet, UK. pp81-86. 
Fullerton J K, Leberl F and Marque R E, 1986. Opposite 
side SAR image processing for stereo viewing. 
Photogrammetric Engineering and Remote Sensing, 
52(9):1487-1498. 
Leberl F, 1990. Radargrammetric Image Processing. Artech 
House Inc, Norwood USA. 
Otto G P and Chau T K W, 1989. A region growing 
algorithm for the matching of terrain images. Image and 
Vision Computing, 7(2):83-93. 
Rocca F and Prati C, 1992. Innovative applications of 
repeated satellite SAR surveys. Satellite Symposium No2, 
ISY Conference Muinich, March 1992. 
Schreier G, Kosmann D and Roth A. 1990. Design aspects of 
a system for geocoding satellite SAR images. ISPRS Journal 
of Photogrammetry and Remote Sensing, 45:1-16. 
Thomas J, Kober W and Leber] F, 1991. Multiple image SAR 
shape from shading. Photogrammetric Engineering and 
Remote Sensing 57(1):51-59. 
Zemmerly M J A, Holden M, Muller J-P, Boffey J, 1991. 
Parallel stereo matching of SPOT satellite images. Digest of 
Colloquium "Parallel architectures for image processing 
applications”, London, 22 April 1991. 
ACKNOWLEDGEMENTS AND NOTES* 
Mia Denos is a former student of UCL sponsored by IBM. 
Christine Clark is also a former student of UCL sponsored by 
SERC and GEC Marconi Research Laboratories. The work 
on the SAR workstation is sponsored by SERC.