In figure 7 the coherence map that is produced 
using the above-mentioned process with the 
artificial data is given. It is clear that the influence 
on the neighboring elements of the point with the 
high backscatter in the center of the images 
disappeared and that the more realistic coherence 
of these neighbors remains. The element in the 
center however kept its high coherence value. 
  
Figure 7 The improved coherence map 
2. Case study 
The procedure described above is used to process 
a tandem pair of ERS-SAR images. These images 
are from the island of Ameland in the northern part 
of the Netherlands. A small part of the images is 
taken of which the landuse and the Ground-cover 
are known and in which region points with high 
backscattering are located. 
2.1 Data description 
The ERS-1 image is taken on 13 June 1996 and the 
ERS-1 image on 14 June 1996. At these days the 
meteorological circumstances were equal, sunny 
and no rainfall in the days before the data 
acquisition. 
The baseline of the two images was 44 meter in 
horizontal and 76 meter in vertical direction. The 
dry weather and the small baseline are ideal to get 
high coherency in those areas where the 
groundcover does not change rapidly. And if low 
coherency appears in some areas in the map then 
that is caused by the changing structure of 
vegetation due to wind. The level of change that 
can be extracted from the coherence map can be 
used as a feature in classifications. ; 
The ground truth of the area of interest is collected 
from the TOP1Ovector databas created by the 
Dutch topographic service. The area that is selected 
is about 2*3 kilometer and lies in the center of 
Ameland. It consists of forest, dunes, a village, 
  
Figure 8 Coherence map of a small area at 
Ameland created with convential filtering 
beach, swallow water and grass. In this experiment, 
however, the area is not classified based on the 
coherence image. Only the influence of the high 
backscatterers is reduced. 
2.2. Coherence map creation 
The co-registration of the Single Look Complex 
SAR images is performed with the INSAR 
processor of Atlantis Scientific. The master (ERS- 
1) and the slave (ERS-2) images are, after the co- 
registration, further processed by in house 
developed software modules. The software modules 
are written in the languages IDL for 1/0 and 
visualization and in C for the filtering and other 
time consuming operations. Data conversion is 
executed in the ITC image processing software, 
ILWIS. 
The dimensions of the filters that are used for the 
  
Figure 9 Coherence map created with a 
combination of filters 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998