| as 
ike 
| of 
ter, 
ced 
ors 
be 
ing 
sor 
INPUT PROCESSING OUTPUT 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
Satellite 
Scanner 9-| Preprocessing 
Data 
I 
Control denen 
Points 9| Mosaicking and 
ot Rectification 
without| with 
DTM DTM 
] 
Radiometrical 
Mosaicking 
] 
Merging 
Multisensor 
Data 
I 
Postprocessing 
I 
Topographic p.| Cartographical 
Information Processing 
I 
Digital Printing 
Screening Original 
  
  
  
  
  
  
Fig. 1: System overview for digital image 
map production 
The production line for the preparation of image maps 
can be subdivided into several processing steps. The 
main topics of each subsystem will be described in the 
following chapters. 
22 Preprocessing 
The preprocessing of the raw satellite image data is a 
prerequisite for all following processing steps and 
influences the quality of the final results remarkable. 
The preprocessing of the data depends on the used 
sensor system. Therefore a wide range of reliable 
algorithms must be available to enhance the sensor 
depending effects in digital images: 
- In the first step all data have to be treated by simple 
histogram stretching in order to make full use of 
the available gray value interval. In the most cases 
the parameters have to be determined for each 
scene individually because they are influenced by 
specific illumination and athmospheric conditions. 
- Sometimes bad pixels or lines caused e.g. by trans- 
mitting errors may arise in digital image data. Those 
errors can be detected and eliminated with a 
method based on thresholding and filtering. 
- All operational satellite scanners show more or less 
typical striping effects depending on the sensor 
construction. If this striping effect is systematically it 
can be eliminated easily by adjusting each line 
integral histogram onto an artificial median integral 
histogram of all lines considered. 
- Especially in dark areas of SPOT data noise influ- 
ences the image quality. This effect can be elimi- 
nated by filtering techniques in the frequeny 
domain. 
689 
- Clouds or dust (particulary in industrial zones) can 
reduce the image quality essentially. This effect can 
be reduced by the treatment of data with partial 
substitution (KÄHLER 1989). 
- From analogue photographs the effect of the 
illuminance fall-off is known. This effect can be 
observed in KFA-1000 images also. Therefore in the 
used software system a module is integrated which 
allows the elimination of illuminance fall-off in 
scanned photographs. 
2.3 Geometrical Mosaicking and Rectification 
It is essential for the integration of image data into a 
GIS to mosaic several scenes in order to fit the re- 
quirement of free map sheet administration of the 
data. The mosaicking must be carried out in a way that 
geometrical and radiometrical differences between the 
images disappear. 
For this purpose a method is used which takes advan- 
tage of the double informations of adjacent scenes. The 
input information is achieved by measuring a few 
ground control points (GCP) in all scenens and tie 
points in overlapping regions. All transformation 
parameters for all scenes of the mosaic are then 
calculated simultaneously in a least squares adjust- 
ment. 
  
a Ground Control Point 
* Tie Point 
m Map Surface 
Fig. 2: Arrangement of ground control points and tie 
points in a satellite image mosaick 
The usage of tie points has also the advantage to 
reduce the number of GCP's. This is very useful 
because it is often very time-consuming to measure 
GCP's in difficult areas (deserts, mountaineous 
regions, etc.) The identification of tie points is done by 
digital image correlation. 
The rectification needs the process of resampling. For 
this different methods can be used. It has to be chosen 
that algorithm which guarantees good preserving of 
the image quality at reasonable computer processing 
time. 
In mountaineous regions the influences of topo- 
graphical relief causes displacements of the pixels 
during rectification. Therefore the polynomials