me
r
*
,
tion S
), P.O
5 Oct
Abstract
only implicit in classification. Fisher et al. (2004) component) can help to identify fuzzy objects.
International Journal of Applied
and Geoinformation 6 (200show that landform objects have a fuzzy nature. A Cheng and Molenaar (2001) proposed a fuzzy
analysis of dynamic coastal landforms, classifying
beach, foreshore and dune area as fuzzy objects. Some
classification errors, however, may occur as only
* Corresponding author. Tel.: +31 53 4874256.
E-mail address: arko@lucieer.net (A. Lucieer).
0303-2434/$ – see front matter # 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.jag.2004.10.0081. Introduction
Object-oriented approaches to remotely sensed
image processing have become popular with the
growing amount of high-resolution satellite and
airborne imagery. Segmentation extracts spatial
objects from an image (Gorte and Stein, 1998;
Lucieer and Stein, 2002). It extends classification,
as spatial contiguity is an explicit goal, whereas it is
straightforward approach to identify fuzzy objects is
to apply a fuzzy c-means (FCM) classification. This
classifier gives membership values of belonging to a
class. The main shortcoming of pixel-based
approaches, such as a standard FCM classifier, is that
these methods do not take into account spatial
relations between pixel values, also known as pattern
or texture. We argue that a texture-based approach
applying segmentation (i.e. including the spatialIn this study, we implement and apply a region growing segmentation procedure based on texture to extract spatial landform
objects from a light detection and ranging (LiDAR) digital surface model (DSM). The local binary pattern (LBP) operator,
modeling texture, is integrated into a region growing segmentation algorithm to identify landform objects. We apply a multi-
scale LBP operator to describe texture at different scales. The paper is illustrated with a case study that involves segmentation of
coastal landform objects using a LiDAR DSM of a coastal area in the UK. Landform objects can be identified with the
combination of a multi-scale texture measure and a region growing segmentation. We show that meaningful coastal landform
objects can be extracted with this algorithm. Uncertainty values provide useful information on transition zones or fuzzy
boundaries between objects.
# 2004 Elsevier B.V. All rights reserved.
Keywords: Multi-scale texture; Region growing; Landform objects; Local binary pattern (LBP) operatorTexture-based landform seg
Arko Luciee
International Institute for Geo-Informa
Department of Earth Observation Science (EOS
Accepted 1ntation of LiDAR imagery
Alfred Stein
cience and Earth Observation (ITC),
. Box 6, 7500 AA Enschede, The Netherlands
ober 2004
www.elsevier.com/locate/jag
Earth Observation
5) 261–270

from a DSM. To identify these objects we argue that
Earth Observation and Geoinformation 6 (2005) 261–270elevation was used as diagnostic information. For
example, an area of low elevation behind the
foredune is classified as beach, whereas it is almost
certainly an area of wind-blown sand removal. Such
errors might be reduced by using spatial or
contextual information, i.e. by considering morpho-
metry or landforms. Cheng et al. (2002) and Fisher
et al. (2004) propose a multi-scale analysis for
allocating fuzzy memberships to morphometric
classes. This can be used to model objects that are
vague for scale reasons. Although this analysis fails
to identify positions of dunes, it is possible to identify
dune ridges and slacks and to monitor their changing
positions.
Regions with a similar reflection can easily be
identified as objects on a remote sensing image. In
case of a digital surface model (DSM) we can use
elevation similarity as a criterion to identify landform
objects. These objects, however, are often character-
ized by more than just elevation. A dune ridge, for
example, has a characteristic profile and/or shape,
which shows a unique texture in a DSM. Therefore, we
argue that texture is an important property of landform
objects and should therefore be taken into account in
landform analysis. We define texture as a pattern or
characteristic spatial variability of pixels over a
region. The pattern may be repeated exactly, or as a
set of small variations, possibly as a function of
position. There is also a random aspect to texture,
because size, shape, color and orientation of pattern
elements (sometimes called textons) can vary over the
region.
Texture measures are used to quantify texture. We
split texture measures into structural (transform-
based), statistical and combination approaches.
Well-known structural approaches are the Fourier
and wavelet transform. Several measures can be used
to describe these transforms, such as entropy, energy
and inertia (Nixon and Aguado, 2002). A well known
statistical approach is the grey level co-occurrence
matrix (GLCM) (Haralick et al., 1973) containing
elements that are counts of the number of pixel pairs
for specific brightness levels. Other texture descriptors
are Markov random fields (GMRF), Gabor filter,
fractals and wavelet models. A comparative study of
texture classification is given in Randen and Husøy
(1999). They conclude that a direction for future
A. Lucieer, A. Stein / International Journal of Applied262research is the development of powerful texturetexture is of utmost importance. A description of
texture reflects the spatial structure of elevation and
slopes, and is therefore indispensable in segmenting
an area into sensible landform units. We start by
modeling texture using the LBP operator at different
scales. Then, we form objects by seeded region
growing. We start at the finest pixel level and cluster
pixels to form objects, based on textural homo-
geneity. Growing of objects is stopped if a certain
threshold is exceeded. A similarity measure is used to
determine whether a pixel can be merged with an
object. This measure also provides useful informa-
tion on extensional uncertainty of objects, expressing
uncertainty in their spatial extent. We expect that
pixels in transition zones show higher uncertainty
values than pixels in core areas with homogeneous
textures. To illustrate the use of texture-based
segmentation for identification of landform objects,
we use a LiDAR DSM of a coastal area in northwest
England. This study builds on work of Lucieer and
Stein (2002) and Lucieer et al. (2003) and further
explores the use of multi-scale texture segmentation
to identify landform objects and to quantify their
extensional uncertainty.
2. Methods
2.1. Texture model—the local binary pattern (LBP)
operator
Ojala et al. (2002) derive the local binary pattern
(LBP) operator by defining texture T in a local
neighborhood of a grey scale image as the joint
distribution of grey levels of P image pixels:measures that can be extracted and classified with a
low computational complexity. A relatively new and
simple texture model is the local binary pattern (LBP)
operator (Pietika¨inen et al., 2000; Ojala et al., 2002). It
is a theoretically simple yet efficient approach to grey
scale and rotation invariant texture classification based
on local binary patterns.
In this study, we implement and apply a region
growing algorithm based on textural information
from the LBP operator to extract landform objectsT ¼ tðg
c
; g
0
; ...; g
P1
Þ (1)