acility 
or under 
mage 
on 
both 
in 
2) 
hieved 
nominal 
width) 
.O m 
onal 
slant 
ata of 
^ to 
1.5 m 
quality 
/ out the 
sets 
fficients 
, Was 
‘ound 
data 
/po- 
m based 
map. In 
r trans- 
ate 
te 
fficients 
n both 
a sets. 
ners) 
ughout 
ions 
ss 709 
upon 
e 
on 
| ves 
sion 
).99940 
a scan 
aan 
ich 
Eh 
Neighborhood processing allows spatial 
processing to be done directly in the 
image domain (Refs.6-7). The underlying 
image algebra allows digital processing 
algorithms to be expressed in terms of 
probing images with other images 
(structuring elements) in a trans- 
lationally invariant way. In particular, 
neighborhood processing can be used to 
reduce speckle noise while preserving 
field boundaries. Isotropic filtering 
consists of a sequence of dilations and 
erosions using structuring elements 
shaped like disks. Disks of small radius 
are employed to clean up noise, disks 
) of larger radii increasingly eliminate 
HH deviations from the average tone pro- 
viding a local approximation to the 
mean value of individual agricultural 
fields. Figure 4 shows the result of 
applying this filtering procedure to 
data set 709 to produce local tone. 
    
Figure 3a. SAR Imagery From 
Data Set 709 (X 
  
Figure 3b. SAR Imagery From 
Data Set 710 Ou) 
  
Figure 4. Local Tone Image From 
Data Set 709 Ou) 
When local tone is subtracted from the 
original SAR image, an image ist ob- 
tained which contains only texture 
(Ref.8). Adding a bias value of 128 
enables both positive and negative 
deviations from the local tone to be 
easily viewed on a digital display. The 
"texture" image obtained from data set 
709 Ou) by this procedure is shown in 
Figure 5. In a sense, this image depicts 
the information which would be discarded 
if analysis was limited to the local 
tone image. 
  
Figure 3c. SAR Imagery From 
Data Set 711 (Xi) 
233 
rm s m— —— 
mer d EEE 
ec re a CNE A RA NEHER Da RD Du LR RE RE -—— eu