T. Object lifting and Image Pyramid 
Object lifting (Kaldun 1988, Müller 1990) is a proce- 
dure, which delivers start values for the heights of 
one Z-facet. The start values are varied and one ite- 
ration is carried out for each start value. Very often, 
there will be a minimum standard deviation of unit 
weight s, with one of the chosen start values and a 
change in the sign of the mean height differences 
after this single iteration. This value is a suitable start 
value for FAST Vision. Object lifting only works, if 
there is such a minimum standard deviation of unit 
weight and such a change of sign, which is not the 
case in areas of low grey value contrasts Cv. fig.1.1) 
  
  
  
  
Ss 
13 
12 
11 
10 
9 
8 
1 
6 
5 
4 
dandi ei inb 
3 200 210 220 ztm] 
Sg 
13 
12 
11 
10 
9 
8 
1 
6 
5 
4 
3 ba ee aaa be 
200 210 220 Z(m] 
equal start values Z for the one Z-Facet onoriginal 
resolution 
Fig. 7.1 so just after one iteration for different start 
values of heights Cabove: good contrasts in the 
images, below: bad contrasts) 
This difficulty should not occur, if object lifting is ap- 
plied in combination with the image pyramid. Here, 
object lifting is used only to determine the start va- 
lues for the heights on the highest level of the pyra- 
mid. If there is no minimum of the standard deviati- 
on of unit weight and no change of sign of the 
mean height differences even on this level, it means 
that the images do not contain coarse textures. In that 
case, simple matching of the images would be diffi- 
cult and of course the application of FAST Vision, too. 
Fig. 7.2 shows the results of object lifting for one Z- 
facet on the highest level (level 2) of the image 
pyramid of the three Dransfeld pictures. There is a 
minimum for the standard deviation of unit weight, 
although its situation cannot be located very clearly 
Cv. fig. 7.28). There is also a distinct change in the 
sign of the mean height differences after that one 
iteration. A good choice for the start value of that 
facet seems to lie between 21] m and 213 m. As it 
can be seen in fig. 7.3, there is only a difference in 
344 
  
the number of iterations on level 2 of the image 
pyramid for the choice of 211 m, 212 m and 213 m. 
Thus, the conclusion can be drawn, that a choice of 
a start value in the vicinity of the minimum of the 
standard deviation of unit weight is sufficient for 
convergence of FAST Vision. 
  
  
  
  
200 205 210 215 220 
Zim] 
equal start values Z for the one Z-Facet onlevel 2 
of the image pyramid 
Fig. 7.2a: sg dependent on start values Z 
Cone iteration carried out for the one facet 
on level 2 of the image pyramid) 
mean difference of heights 
after one iteration 
  
T T T T T 
  
  
  
  
Si : 
ok : 
-5L ; : 
1 1 1 1 L 
200 205 210 215 220 
Z{m] 
equal start values Z for the one Z-Facet onlevel 2 
of the image pyramid 
Fig. 7.2b: mean difference of height after 
one iteration dependent on start value Z 
(one iteration carried out for the one facet 
on level 2 of the image pyramid) 
  
QD) v" t= tf 4 C ul 3 
£e 
N n rr rt o0 
  
  
Fig. 
for