International Archive 
landscape. The procedure was aimed at the extraction. of 
s of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, P 
art B4. Istanbul 2004 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
! Specification heavily depends on system resources, particularly m 
4 . ; . > > ; 
convertible from divers raster formats, e.g. GeoTIFF, IMG; 
image size is modified; 
Cognition elements) is available s 
conversion was not possible; 7 State April 2003; * A simplified version (e 
Table 1. Outline about evaluated segmentation software. 
relevant land cover/use object boundaries. The segmentations 
were produced by the developers — except for eCognition, 
SPRING and the Erdas Imagine extension ‚Image Segmen- 
ain memory; > for the used imagery (2000 by 2000 Pixel); 
? Proprietary vector format; ? Internal vector 
ince May 2003 (3500€) 
Segmentation eCognition Data CAESAR 3.1 | InfoPACK Image Seg- Minimum SPRING 4.0 | 
program 2.1 resp. 3.0 Dissection 1.0 mentation Entropy 
Tools (for Erdas Approach 
Imagine) 
Developer Definiens INCITE, N.A.Software | InfoSAR Ltd. | USDA Forest | University of National | 
Imaging Stirling Ltd. Service, Re- | Bonn, Inst. of | Institute for 
GmbH University mote Sensing | Computer Space 
Applications Science Research 
Center 
Website WWW. www.incite. WWW. WWW. www2. www-dbv. www.dpi. 
definiens- org.uk/ nasoftware. infosar. erdas.com/ informatik. inpe.br/ 
2 imaging.com projects/ co.uk co.uk SupportSite/ | uni-bonn.de/ spring 
s polyDemo/ 
E | Algorithm Region Superpara- Simulated Simulated Region Triangulation Region 
= growing magnetic annealing annealing growing growing/ 
a clustering Watershed 
Field of Remote Image Remote Remote Remote Polygonizatio Remote 
application sensing analysis, sensing, esp. | sensing, esp. sensing n of noisy sensing 
statistical radar data radar data imagery 
physics 
Fundamental Baatz & Ferber & Cook et al. Cooketal. | Ruefenachtet| Hermes & Bins et al. 
reference Schäpe 2000 Wórgótter 1996 1996 al. 2002 Buhmann 1996 
2000 2001 
State of 04/2002 resp. 10/2002 09/1998 03/2003 02/2002 08/2002 08/2003 
development 11/2002 
Operating Win Win, Linux, Linux, Linux, Win Win Linux Win, Unix 
system Unix, SGI Solaris 
= System Stand-alone | N/a (maybe | Stand-alone | Stand-alone Erdas Stand-alone | Stand-alone 
3 | environment MatLab) Imagine 
8 | Number of 3 3 6 2 2 2 2 
£ | parameters 
= Ca. runtime‘ 10 min 10 min N/a (long) 10 min 1,5 h Several hours 30 min 
| Reproduce- No resp. No No No No No No 
ability? yes (ver. 3.0) 
Classification Yes (Fuzzy No Yes Yes No No Yes 
support Logic, Near. (Maximum 
Neighbour) Likelihood) 
Max. image 10000 by 4000 by 2000 by No 2000 by 2000 by 2000 by 
= size [ca. Pixel]' 10000 4000 2000 Limitations 2000 1500 2000 
£ | Max. bit depth 32 bit 16 bit 16 bit 64 bit 8 bit 8 bit 8 bit 
o Input formats Raster, Shape | Raster (TIFF) IMG NetCDF? IMG Raster GRIB' 
= Vector output Shape No (external | No (external | No (external ArcCoverage | No (external | No( external 
+ | format conversion) conversion) conversion) conversion)” conversion)’ 
™ | Use of external Yes No Yes Yes Yes Yes Ja 
data 
Availability Commercial | Commercial Out of Commercial Freeware On request Freeware 
8 distribution 
3 Ca. costs’ 14.000 €/ N/a resp. - 8.300 € / free - free 
B (commercial/ 9.000 € free (non- 5.800 € 
A | research) (non-profit) profit) 
/2.900 €' ——— 
? when 
tation’. Thus, optimal segmentation results (e.g. parameter 
settings) by experts were assured. 
1098 
Int 
Re 
rur 
Ap 
101 
the 
par 
jeg 
seg 
32 
33 
Firsi 
crite 
mea 
line: 
desc 
focu 
Furt! 
and 
20 € 
conti 
visu 
dant: 
facto 
and ( 
of o 
refer 
AS pi 
refer 
lands 
calcu 
Squar 
Was y 
À ge 
differ 
the € 
Furth 
Seem.