3. CARTOGRAPHICAL DATA PROCESSING 
The entirely digital approach to Satellite Image Map 
production also includes the digital generation, place- 
ment and combination of cartographic elements, i.e. 
names, symbols, coordinates etc. The integration of 
graphical elements into image data is a technical pro- 
blem which can be handled by interactive operations if 
the relevant software components are available. 
Nevertheless, this task is also a new problem for carto- 
graphical design. Cartographers have a long experience 
as far as the combination of graphical elements is con- 
cerned. But so far, there is only very little experience 
with the integration of letters, lines, symbols, etc. into 
image data. 
These graphical elements can obviously be integrated 
into the image map by simply printing it in black. 
However, very often the image background is dark 
and therefore black elements can hardly be recognized. 
On the other hand they can be integrated in a negative 
form, which is also easy to achieve. But it es evident, 
that similar problems occur in light areas of the image 
and that the graphical data can not be recognized. 
In principle it is an old problem of cartography to in- 
tegrate letters, symbols etc. in such a way into the back- 
ground of the map, that they can easily be recognized 
and the map content is disturbed as less as possible. It 
is a quite common technique, to remove the surroun- 
ding parts of the background for this purpose. But this 
approach, although very successful in topographic 
mapping, does not yield satisfying results in the case of 
image maps. 
Therefore a number of related experiments has been 
carried out. The purpose was, to find a solution for 
this problem by making use of the very flexible digital 
image processing techniques. The preliminary results 
of these experiments are shown in the Figures 3 to 8. It 
goes without saying, that the black and white re- 
production can only partly transfer the impression of 
the colored images. 
The original image data set (Fig. 3) is part of a Satellite 
Image Map (Laag, Somalia, 1:50,000) in the colors red, 
green and blue, as it is usual in digital image pro- 
cessing. Fig. 4 shows the graphical elements to be 
integrated in these image data. If these elements are 
simply printed in black, this results in Fig. 5; within 
the dark areas of the image the lines and letters are 
more or less invisible. The alternative, namely to inte- 
grate the graphical data in negative form by removing 
it from the image data, does also not produce a satisfy- 
ing result. Fig. 6 shows, that the graphical elements are 
difficult to recognize in the light areas of the image. 
Furthermore, the white elements are highly domi- 
nating, and therefore the overall impression of the 
image is still more affected than in the previous case. 
Attempts to solve the problem should obviously re- 
tain the graphical elements in black, but improve its 
recognizability in dark areas. This can be achieved by 
applying digital image processing techniques. Some 
tentative studies made clear, that the highest fexibility 
for this task is achieved after the derivation of the 
black color data set out of the red, green and blue data. 
312 
In most cases modifications of this black channel can 
solve the problem, so that the basic image can remain 
untouched. The most successful approach, which has 
been tested so far, is the following: 
- The graphical data are converted to a raster format 
which is identical to the image data. 
— By means of blurring filters an unsharp mask of the 
graphical elements is generated. 
— The unsharp mask is inverted and added to the 
black channel data; thus the areas of the graphical 
elements become lighter and this effect blurs in 
their direct neighbourhood. 
— After that the graphical elements are added to the 
data of the black channel. 
The result of this procedures is shown in Fig. 7. The 
recognizability of the graphical elements in the dark 
areas is substantially improved. However, the opera- 
tions also work in the light areas of the image. Thus 
some halo effects around the elements occur at places 
where this is not necessary for their recognition and is 
degrading the image to a certain extent. In order to 
avoid this undesired results the operations can be 
modified in such a way, that only the darker parts of 
the image are affected. This leads to Fig. 8, showing a 
good recognizability of the graphical elements and a 
minimum of degradation of the image itself. 
Further experiments will be necessary to optimize the 
parameters of this approach and also to find out the 
appropriate letter types and symbols for this task. 
4. SATELLITE IMAGE MAPS 
FROM PHOTOGRAPHIC DATA 
The concept described above was designed for the use 
of digital satellite image data as they are acquired by 
opto-mechanical or opto-electronical scanning sy- 
stems. However, the software package can also be 
applied to digitized data from photographic cameras. 
A related experiment was carried out in cooperation 
with the company KAZ Bildmess GmbH in Leipzig. 
Photographic data from the Russian camera KFA-1000 
were used to generate a Satellite Image Map from the 
city of Leipzig and the surrounding area in the scale 
1:50,000. The original photograph was scanned with a 
resolution of 12.5 um by means of a HELL scanner. 
Because of the fact that the color film uses two layers, 
only two sets of digital data were really useful, the 
third channel was more or less noise. However, in 
order to make a third data file available for further 
processing, an 'artificial' channel was derived by mo- 
difying the red data set through an empirically defined 
look-up table. The result of this procedure was used as 
the blue channel in further operations. An additional 
correction for the illuminance fall-off had to be 
applied. Otherwise processing followed the procedures 
described above. 
The result of the test was an image map of similar 
quality as it can be achieved from SPOT/TM-com- 
binations. Some details, e.g. the texture of forest areas, 
were reproduced better, on the other hand photo- 
graphic granularity became visible and the differen- 
tiation of colors was reduced. Alltogether the experi- 
ment was very successful.