tanbul 2004 
  
  
> Catalunya 
> of Digital 
XS data on 
4 (red) 
aggam __ 
x 1411pix 
ing: 5m 
x 141 1pix 
ing: Sm 
x 141 1pix 
ing: Sm 
x 1411pix 
ing: 5m 
x 9201 pix 
ing: 10 m 
  
ter values) 
e. All the 
the size of 
g between 
have been 
> ones the 
the invalid 
ving ones : 
d standard 
dots". 
stics we've 
se most of 
ffects with 
  
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B1. Istanbul 2004 
significant values like 300 m or even more (see examples in 
the next pages). 
All this will be the same for Bavaria site. 
2.4 Independent results 
For the global area and the area6, we took off (mask) as well 
the sea zero value, to have more significant results., presented 
in Tables 4, 5 and 6. 
. Kornus min max mean st. Dev 
area? 15 0,8 3,8 
area4 32 1,2 6,5 
area5 14 17 3,6 
area6 27 11.0 7.0 
Table 4: Results from W. Kornus DEM. 
  
  
P. Reinartz min max mean st. Dev 
  
  
  
  
  
  
  
global -6 28 10,3 3.3 
  
Table 5: Results from P. Reinartz DEM. 
H. am min | max mean 
areal 33 0,1 
area 16 I 
area2 HRG/HRS 22 3,4 
area4 107 0,6 
area4 HRG/HRS 28 4,4 
area6 27 8,8 
area6 HRG/HRS 32 10,9 
lobal 27 0,1 
Table 6: Results from H. Raggam DEM. 
  
All the results (apart from H. Raggam DEM on area4) show 
standard deviation lower than 10 m even in high relief areas. 
The P. Reinartz DEM seems to have an altitude bias with the 
reference (nearly 10 m), in his document P. Reinartz explains 
that this result can be improved with taking into account 
ground control points (Reinartz, 2004). 
We can mention.as well a significant bias in area6 for all 
DEM. 
2.5 Difference images analysis 
Difference images are presented in Fig. 7, 8 and 9, in which 
all red values describe positive values, that means that the 
received DEM is higher than the reference. In the conversly 
opposit, all blue values describe negative values, that means 
that the received DEM is lower than the reference. We 
change the LUT (Look Up Table) for the first difference 
because of the noticed bias. 
   
       
    
Fig.8 H. Raggam DEM - Reference difference 
Fig.9 P. Reinartz DEM — H. Raggam DEM 
    
= 
  
  
Fig.7 P. Reinartz DEM — Reference difference ( - bias 10m) 
  
Bo-: 
2 
A look on the difference images (especially with global 
DEM) shows different things: 
the bias observed on P. Reinartz DEM is constant (more 
or less constant blue) apart from an area near the sea and 
closed to area6. 
This bias (area 6) is visible as well on difference between 
H. Raggam and reference but not in the difference 
between the P. Reinartz DEM and the H. Raggam DEM, 
in fact this area corresponds to the city. Space DEM are 
higher because the altimetric restitution follows more or 
less the top of the buildings (or forests) which is 
obviously not the case for the reference (the reference is 
perfectly flat on this area). 
There is no visible bias in statistics for the H. Raggam 
DEM but in fact we can see a north/south low frequency 
waveline distorsion, the difference is negative in upper 
north, positive after and again negative in the south. 
Those are small difference (1 or 2m, 3m max in morth) 
which must be caused by modelisation of the HRS 
camera.