667 - 
STATISTICAL EVALUATION OF MULTISPECTRAL SCANNER IMAGES 
E. Triendl. G. Kritikos, M. Schroeder, B. Sahai 5: ) 
DFVLR-Forschungszentrum Oberpfaffenhofen 
D 8o31 Oberpfaffenhofen 
Abstract 
An important method for classification of multispectral scanner 
images is the ratio-mapping method. The classification of objects 
whose signatures vary from time to time, from climate to climate 
and from season to season proves to be quite a difficult task. To 
overcome some of the difficulties, the ratio of grey levels of the 
same picture element in two different spectral bands is determined. 
A frequency distribution of these ratios as well as the variance T 
is obtained. Once is determined, the feature extraction task is 
very much simplified. The accuracy of the method is further im 
proved by dividing the picture in segments and applying the above 
procedure to each segment. Making use of these procedures a classi 
fication map of the Bavarian landscape has been attempted. 
) guest scientist from the Space Applications Center, 
Ahmedabad-15, INDIA