A 
If the texture region is composed by different texture elements 
the structural textüre analysis generates. (10) different region /6/ À 
boundaries. A statistical texture analysis in the defined 
segments indicated which texture element is texture primitiv 
and the complex textured object (15) boundaries are detected 
with high resolution. : 
c) If the image is composed by both, spectral homogeneous and 
textured objects the cases a) and b) occur together. The /1/ 4 
information extracted by the different steps will support the 8 
other steps. Large textured areas (8) for example can be S 
surrounded by spectral objects and need no further F 
investigations. 
/8/ P 
F 
G 
CONCLUSIONS 
The preliminary results of feature combination are promising. The "his r 
informations extracted by one proceeding step are able to support German 
the other ones. However the proposed algorithm has to demonstrate DFVLR 
this efficiency on different types of image data. The use of 
parallel, pipelined or special purpose computer systems will not 
only reduce the computation time, but will also allow the 
investigation of techniques which may otherwise be considered 
infeasible. 
REFERENCES 
/1/ Kazmierezak, H. Erfassung und maschinelle Verarbeitung 
von Bilddaten. 
Springer Verlag, 1980, Berlin. 
/2/] Mauer, B. Untersuchung zur Auswertung mehrkanaliger 
Sehürf, R. Information mit lokalem Kontext 
FIM-Bericht Nr. 94, Karlsruhe, 1081. 
/3/. Baraët, m. Rechnergestützte Ermittlung und Bewertung 
von Texturparametern aus Bilddaten fiir die 
Objektklassifizierung in der Fernerkundung. 
DFG-Bericht Ka 414/3, 1981. 
4/7 Schaerf, R. Untersuchungen zur bildgesteuerten Separie- 
rung von Objekten in multispektralen Bild- 
daten. 
Dissertation, Fak. flir Elektrotechnik, 
Universitat Karlsruhe, 1981. 
/5/ Rosenfeld, A. Texture Synthesis 
Tipkin, ‘B.S. in: Picture Processing and Psychopictorics 
Lipkin, B.8., Rosenfeld, A. (Hrag) 
Academic Press, New York, 1970, pp. 509-545 
42