Stephane Paquis
(a) Surface Dressing (b) Porous Asphalt
(c) Bituminous Asphalt (d) Ultra Thin BA
Figure 1: Pavement surface families
filters. These morphological filters are used in section 3 to compute an original morphological pyramid. The following
section presents the multiresolution cooccurrence matrix and its analysis to obtain a texture description. Section 5 gives
classification procedure and some results.
2 MORPHOLOGICAL TOOLS
Mathematical morphology analyses the geometrical structure of a set by probing and transforming its microstructure with
different predefined elementary sets (Toet, 1989), also called structuring elements. The two fundamental operations are
erosion and dilatation. Dilatation removes low intensity regions. À function f(x), dilated by a structuring element B, is
defined as :
Vee DC E*,. (f 6 Byz) — max(f(z y)). (1
y
where & denotes dilatation, E is the set of integers and B a subset of E?. Similarly, an erosion removes small regions of
high intensity, and is defined as :
Va € DCE’, (f© B)(z) = min{f(z — y)}, 2)
y
where © denotes erosion.
Morphological filters are used to build morphological pyramid in order to reduce information content by eliminating small
objects or object protusions from high resolution and to produce a signal convenient for subsampling step (Haralick et al.,
1987). Morphological filters are constructed by iterative application of erosion and dilatation. The most frequently used
is the opening one, which consists of an erosion followed by a dilatation, defined as :
(foB)-(foB)aB, (3)
where o denotes the opening operation. In a general way, an operator 7 is defined as an opening if and only if the next 3
rules are checked (Serra, 1988) :
686 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000.
Figure
by sha
The la
filter o
As one
size of
structu
multid
3 M
The m
texture
usual «
on the
An im
course
pyram
where
Equati
accour
is link
develo
À reve
2Nx2,
Operat
and id:
related