X fitering (1, j) = G(i, j)* x(i, j) (® 
where y is a variance and it decides the 
filtering characteristic. (Fig.9) 
Figure 10 shows the analyzed contours using 30 
dB images which are applied to filtering methods. 
Figure 10(a) is the output of the narrow band 
filter against the noisy image. The narrow band 
filter can reduce the influence of noise, but it 
causes the deterioration of the edge characteristic 
at the top of mountain. On the other hand, in 
Fig.10(b) the wide band filter does not cause the 
appears at the foot of mountain. 
  
(b) S/N=40dB 
Fig. 7 Disturbance by noise 
without filtering. 
    
  
     
[7 
A LA N 
HT AR 
IGN 
INN 
DLR 
í ESS 
ESS 
     
  
     
    
  
25 
20 
3 | ^» Top area 
5 10 
5 
0 
(b) v 20.5 
  
60 50 30 20 
40 
S/N(dB) 
  
  
  
Fig.8 Restoring errors the top and foot area. 
2-3.LOW PASS FILTER 
Generally, gaussian filter is used as low pass 
filter. The gaussian function is 
  
= (Un) axp(-(1* +j2)/2y?) .(3) (a)Using narrow band filter (y — 1.0) 
And, one obtains to a smoothed image according 
to 
312 Intemational Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998 
edge characteristic deterioration, but a little error 
  
Fig 
2-4.Al 
Here 
restra 
chara 
autom 
to geo 
simpk 
Figure 
The p 
level's 
differe 
filter : 
band 1 
If 
IS 
If 
Fig.11 
The 
applie 
measu 
to this