a) High-pass filtered input image of a sample area of spruce, 
b) binary image derived from it by hard-clipping. 
Analysis techniques and classification results 
The functional relationship between hardclipping threshold density and 
the three primary textural parameters are shown in Fig. 5 for typical 
examples of the four tree species spruce, fir, pine and beech. 
Although these diagrams have been obtained from measurements at 
individual sample areas, they are typical for the tree species under 
observation. The functional relationships will change, however, with 
a change of the angle of view under which the sample area has been 
photographed. 
Many of the characteristic features of the curves in these diagrams 
can easily be traced back to characteristic features of the various 
tree species: As an example, it can be seen from a comparison of 
the diagrams of spruce and pine in Fig. 5, that the ratio perimeter/ 
area is smaller for pine than for spruce. This is explained by the 
fact that, on aerial photographs, spruce is characterized by a star- 
like shape, while pine has a more compact structure when viewed 
from the air. 
From these diagrams ''secondary textural parameters' were derived 
in the following way: Those threshold densities were looked up at 
which the area is 90% (a) and 10% (c) of the total area, respectively. 
The threshold density b was defined as the arithmetical mean of 
a and c. All further evaluations were then performed on the basis of 
these three threshold densities a, b and c. As a consequence, the 
extracted textural parameters are practically independent of the 
absolute density values, so that, e.g., variations in the