45 
Row- and 
Column- 
Doubling 
Rectification and 
Resampling with 
Bilinear Inter 
polation 
Classifi 
cation 
results of doubled 
and geometrically 
:ied Data 
Doubled 
2d 
4.03 
9.02 
8.25 
28.54 
34.90 
15.04 
I (%) 
lich disappear in 
result of doubled 
ling of data before 
ment of classifica- 
lds of data prepro- 
the results. The 
ns are briefly dis- 
influences of the 
rast enhanced data 
: linear histogram 
;d out separately, 
neighbourhood re- 
ange in respect to 
is is valid for re- 
nhanced data. 
interpolated data 
t to the result of 
l occurs in isolated 
classes with a lot 
of highly structured areas show the greatest chang 
ing. The use of contrast enhanced data for rectifica 
tion and resampling instead of original data seems to 
show a slightly better accuracy. 
Classification of resampled data by bilinear interpo 
lation may be improved if a row- and column 
doubling before rectifying is carried out. 
The above experiences indicate that contrast enhancement 
influences classification results slightly, either if data are 
geometrically preprocessed or not. Moreover classification 
of nearest neighbour resampled data leads to a result, 
which is only little influenced by preprocessing, even if a 
great geometric transformation (from satellite image to 
map grid) is performed. 
If bilinear interpolation for resampling is used the classi 
fication results will tend to more differences in respect to 
those of the original data. 
Bilinear interpolated data may be used for classification 
if a doubling of data before rectifying is carried out. The 
improvement of classification result is obvious, but study 
ing of classification accuracy needs more detailed investi 
gations. 
In the whole, the investigations dealed in this paper show 
the influences of various preprocessing methods. Therefore 
the purposes for the digital classification of different kinds 
of image preprocessing must be considered carefully. 
REFERENCES 
Forster, B.C., Trinder,3.C., 1984. Effects of Resampling of 
Landsat Data on Classification Accuracy. Aust. Л. Geod. 
Photo. Surv., 40, Липе 84: 53-67 
Kähler, M., Milkus, I., 1986. Berlin from Space - A Digital 
ly Produced Satellite Image Map. Intern. Symp. on 
Mapping from Modern Imagery, Edingburgh September 
1986 (in print) 
Realnutzungskartierung und Flächenbilanzierung, Maßstab 
1:50000, im Bereich des Raumordnungsverbandes Rhein- 
Neckar. ifp-Institut für Planungsdaten und StadtBauPlan 
Frankfurt am Main/Darmstadt 1976