International Archives of Photogrammetry and Remote Sensing, Vol. 32, Part 7-4-3 W6, Valladolid, Spain, 3-4 June, 1999 
28 
Digital elevation model 
Landsat TM, band 2 
SPOT XS, band 1 
Airborne SAR 
JERS-OPS, band 3 
JERS-SAR 
Fig. 9. Geocoded multisensor images and DEM of the mountainous area “Oetztal” (Austrian Alps). 
registration. These results indicate the potential of coregistration 
of already geocoded images to improve their geometric 
accuracy. However, the images have to be similar enough to 
perform reliable image matching. 
Table5. Matching accuracy of geocoded Landsat TM and 
fine-registered SPOT ortho-image. 
East (pel) 
North (pel) 
Distance (pel) 
Mean 
-0.05 
0.00 
0.49 
Std. Dev. 
0.45 
0.44 
0.36 
Figure 11 shows the problems inherent to the coregistration of 
ERS-SAR data originating from ascending and descending 
orbits, especially if acquired over mountainous terrain (here, 
alpine area in Austria). As shown in the figure, the input images 
are extremely different due to the geometric distortion effects 
and radiometric peculiarities induced by the terrain topography 
and the different illumination directions. In case of such 
ascending/descending ERS images, it is difficult, even after 
geocoding, to identify common features, as these are affected by 
extended layover areas spreading over the slopes facing the 
ERS sensor during the ascending and descending overflights. 
It is obvious that in this example geocoding using parametric 
models and a DEM is the only way to register such input image 
data with acceptable accuracy, as they are very different due to 
radiometric as well as geometric sensor characteristics. An 
example of a straightforward superimposition of the 
ascending/descending ERS images is shown in Figure 16. 
Fig. 10. Displacements between geocoded Landsat TM and 
SPOT XS images. 
4.3. Image-to-Image Parametric Registration Example 
Image-to-image registration may be useful in case that the 
radiometric and geometric nature of one of the registered 
images should remain unchanged. An example for the 
transformation of a geocoded image data into the geometry of 
another image is shown in Figure 12. Here, an optical JERS 
image was first geocoded and then a map-to-image 
transformation was performed with regard to a JERS-SAR 
image.