702 
Multilayer composition - grey level reduction 
  
  
  
  
Figure 3: Multi-layer composition. 
and needs to take into account the terrain relief 
changes using a suitable DEM. 
In order to perform the classification, the images 
(coherence, amplitude and SPOT) have been geo- 
coded using the DEMs generated with the SPOT and 
InSAR data described in paragraph 2. 
In the following, a short description of their input 
images and their geocoding is given. 
3.1.1 Coherence Image 
Beside DEMs, coherence maps represent an other 
important product of SAR interferometry. The pair of 
InSAR images usually comes from two different 
passes over the same area. The important assumption 
to perform interferometry (i.e. to compare the phases' 
of the two images) is that there are no changes in the 
terrain between the two passes. The coherence image 
is the cross-correlation of the two images and indi- 
cates, pixel by pixel, how similar are the signals gath- 
ered during the two takes. If changes occur (due, for 
instance, to rainfall), the signals can be very different 
and the phase of the interferogram can become use- 
less for DEMs. 
The coherence is a good indicator of the phase quality 
for DEM generation. Furthermore, as it is related to 
the physical and geometric properties of the terrain, it 
can be exploited to gain information about the earth's 
surface (e.g. for the identification of geological fea- 
tures, for the classification of the land cover, etc.); see 
[Wegmüller and Werner 1996]. 
The coherence image is characterised by the same 
geometry of the original SAR images. During the 
DEM generation it is possible to transform a coher- 
ence image in a coherence map going from the imag- 
ing geometry to a chosen cartographic projection, i.e. 
obtaining the geocoded coherence. 
The image coherence employed for the classification 
has been estimated from the same ascending ERS-1 
images described in paragraph 2.1.1. The geocoding 
has been performed at the same time of the InSAR 
DEM generation. It works pixel wise: for each pixel 
of the interferogram (so for each pixel of the coher- 
ence image) the 3D position of the imaged point has 
Intemational Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998 
bec 
ent 
(wi 
bee 
3.1 
Th 
rive 
eac 
The 
ram 
em 
the 
des 
spe 
abo 
bee 
gen 
cod 
ado 
The 
has 
Pho 
witl 
DE