tanbul 2004 
—— —— 
D hort 
LÀ 
'reas Bare Lands 
inds 
ural Areas 
Yody 
ces 
ex 
ns 
; 
reas+Bare Lands 
nds 
Pine) 
ls 
ural Areas 
jodies 
ces 
Fier map 
A COMPARISON OF SEGMENTATION PROGRAMS FOR HIGH RESOLUTION 
REMOTE SENSING DATA 
G. Meinel, M. Neubert 
Leibniz Institute of Ecological and Regional Development (IOER), Weberplatz 1, D-01217 Dresden, Germany - 
G.Meinel@ioer.de, M.Neubert@ioer.de 
PS ThS9: Uncertainty, Consistency and Accuracy of Data and Imagery 
KEY WORDS: Automation, Land cover, Classification, Software, IKONOS, Performance, Comparison, qualitative 
ABSTRACT: 
Methods of image segmentation become more and more important in the field of remote sensing image analysis — in particular due 
to the increasing spatial resolution of imagery. The most important factor for using segmentation techniques is segmentation quality. 
Thus, a method for evaluating segmentation quality is presented 
and used to compare results of presently available segmentation 
programs. Firstly, an overview of the software used is given. Moreover the quality of the individual segmentation results is evaluated 
based on pan-sharpened multi-spectral IKONOS data. This is done by visual comparison, which is supplemented by a detailed 
investigation using visual interpreted reference areas. Geometrical segment properties are in the focus of this quantitative evaluation. 
The results are assessed and discussed. They show the suitability of the tested programs for segmenting very high resolution 
imagery. 
1. INTRODUCTION 
Segmentation means the grouping of neighbouring pixels into 
regions (or segments) based on similarity criteria (digital 
number, texture). Image objects in remotely sensed imagery are 
often homogenous and can be delineated by segmentation. 
Thus, the number of elements as a basis for a following image 
classification is enormously reduced. The quality of 
classification is directly affected by segmentation quality. 
Hence quality assessment of segmentation is in the focus of this 
evaluation of different presently available segmentation 
software. 
Despite some early research activities (e.g. Kettig & Landgrebe, 
1976), image segmentation was established late in the field of 
remote sensing. First beginning with the availability of very 
high resolution imagery (< 1 m) and their characteristics (high 
level of detail, spectral variance etc.) this method has become 
popular as a common variant of data interpretation. 
Recent investigations have shown that a pixel-based analysis of 
such high resolution imagery has explicit limits. Using 
Scgmentation techniques some problems of pixel-based image 
analysis could be overcome (e.g. Meinel, Neubert & Reder, 
2001). 
This paper is not related to more mathematical surveys of 
Segmentation, like Haralick & Shapiro (1985) or Pal & Pal 
(1993). It is rather a more application-oriented comparison 
based on real remote sensing data. 
Among Segmentation software there is a growing number of 
feature extraction programs. In contrast they do not 
fractionalise the whole image but rather select specific objects 
from imagery. APEX (PCI Geomatics), FeatureXTR (Hitachi 
Software Global Technology) or Feature Analyst (Visual 
Learning Systems) belonging to this group of tools, which are 
Not considered herein, 
2. EVALUATED SEGMENTATION SOFTWARE 
Recently there exists a multitude of implemented segmentation 
algorithms for remote sensing tasks, partially having very 
different characteristics. Only some of them are available 
commercially. Often they are developed by research institutions 
or universities. For evaluating capabilities of different 
algorithms the following programs were compared: 
e  eCognition 2.1 resp. 3.0 (Definiens Imaging GmbH, 
Munich, Germany); 
* Data Dissection Tools (INCITE, Stirling University, 
UK); 
e CAESAR 3.1 (N.A.Software Ltd., Liverpool, UK) 
® InfoPACK 1.0 (InfoSAR Ltd., Liverpool, UK); 
e Image segmentation for Erdas Imagine (USDA Forest 
Service, Remote Sensing Applications Center, Salt 
Lake City, USA); 
e Minimum Entropy Approach to Adaptive Image 
Polygonization (University of Bonn, Institute of 
computer science, Bonn, Germany); 
e SPRING 4.0 (National Institute for Space Research, 
Säo José dos Campos, Brasilia). 
All programs are described in brief in table 1. The choice of 
approaches was based on the software segmentation suitability 
for remote sensing imagery. On the other hand cooperativeness 
of the developers was a precondition for this survey. 
3. METHOD 
3.1 Used imagery 
Pan-sharpened multi-spectral IKONOS data (I m ground 
resolution, principle component algorithm) of two test areas 
were segmented by the software above. Each test area has a size 
of about 2000 by 2000 Pixel, representing an urban and a rural 
1097