ion"
n
THE PROBLEM OF TEXTURE DEFINITION FOR IMAGE ANALYSIS
GEOSCIENCES
by,
Ulrich Wieczorek.
Institut für Geographie der
Universität München, Abtei-
lung Fernerkundung am Lehr-
stuhl Prof. Gierloff-Emden
Abstract
By computing shape parameters of texture a more
exact definition of image structures called "tex-
ture" is possible. From this definition "arrange—
ment parameters" of texture can be derived
"7. Introduction
A. Rosenfeld regards textures as "complex visual patterns, com-
posed of entities or subpatterns" (A. Rosenfeld and A. Kak,
"Digital Picture Processing",New York,San Francisco,London,1976,p.59).
A lot of attempts are made to describe the image textures. It
should be mentioned the remarkable ideas of R.M. Haralick and
the investigations of the research team at the FIM in Karls-
ruhe, Germany. One purpose of these investigations is to melio-
rate the classification methods by supplying the spectral sig-
natures by textural signatures. Not only automatic mapping of
surfaces with different spectral behavior but also automatic
mapping of different surface structures is desired.
2. Shape Parameters of Texture
For the identification of groups of objects on the earth surface
by aid of spectral respectively textural features a correlation
between these features and characteristical properties of the
object groups on the earth surface has to be constructed.
For this purpose ground checks and ground trooth measurements
are necessary. À correlation is possible between some spectral
properties of the objects on the earth surface and the greytone
in the different channels of a more channel image.
A correlation between testural features as used e.g. by Haralick
(2.) and object properties is. not possible, because it is very
difficult to. find measuring parameters on the earth surface which
are equivalent to. the parameters of texture. The features defined
by R.M. Haralick (2 ) cannot be computed for the reflection variance
on the earth surface, because it would be necessary to. measure
the intensity of reflexion in each point of a test site on the
earth surface at the same time, This will be impossible. So other
properties have to be found which cab be measured easilier on the
earth surface. These must be properties which are interesting
for the geoscientist.