A DIGITAL IMAGE PROCESSING APPROACH TO CREATING DTMs FROM DIGITIZED CONTOURS 
Morakot Pilouk 
Research Officer, Royal Forest Department 
National Forest Land Management Division 
Phaholyothin Road, Chatuchak, Bangkok, Thailand 
Klaus Tempfli 
Associate Professor, Department of Geoinformatics 
International Institute for Aerospace Survey and Earth Sciences (ITC) 
P.0.Box 6, 7500 AA Enschede, The Netherlands 
ABSTRACT 
DTMs 
by automatically extracting 
interpolation. A hybrid set 
generation, including distance 
following, interpolation, and 
IBM-compatible PCs and is based on ILWIS. 
Skeleton lines 
transformation, 
KEY WORDS: 
cost-effectiveness 
INTRODUCTION 
"Recycling" graphic contour lines remains popular 
for spatial analysis in a digital environment. The 
availability of topographic maps in many 
countries, combined vith the low decay rate of 
terrain relief data at medium and small scales, 
rapidly decreasing costs of scanners, PCs with 
digitizing tablets and ‘GIS software’, favour 
A/D-converted contours as data source for digital 
terrain relief models (DTM). Compared with a 
direct ground survey or photogrammetric 
techniques, data collection is cheap and does not 
require special expertise. 
Defects of Contours 
a digital surface description that is 
rooted in (graphic) contours suffers from limited 
quality. Non-photogram-metrists are often 
surprised that such a DTM is of lower fidelity 
than expected, especially when using it in erosion 
studies and the like, where slope and break of 
slope are important properties. One reason for the 
limited quality is the unfavourable sampling 
pattern inherent to digital contours. The high 
sampling density within lines and no information 
on terrain relief between these lines implies 
sub-optimal fidelity of digital surface 
However, 
representation. Another reason is the wealth of 
errors in contours depicted in maps. In addition 
to the errors of photogrammetric operations 
(aerial triangulation, orientation, measuring and 
plotting) and deformations of materials used, 
there are even more disturbing errors introduced 
by cartographic operations. Examples include 
displacement of contours when generalizing, "slope 
allowance" of contours in hilly and mountainous 
terrains, replacement of contour lines by symbols 
in areas with cuts and embankments, cliffs, etc. 
Added to these are the errors of the A/D 
conversion [12]. 
Figure 1 shows a typical example of a surface 
obtained from digitized contours and 
straightforward linear interpolation between the 
nearest contour points. Cut off hill tops, filled 
troughs, and terraces along ridge and drainage 
lines are disturbing artifacts for many DTM 
applications. These disturbing effects of common 
interpolation methods, when using bulk data such 
as digital contours, can be reduced/diminished by 
Dirichlet tessellation, thinning, 
establishing topology. The 
956 
derived from digitized contour lines often suffer from limited quality. The quality can be enhanced 
from the 
of raster and vector techniques is 
contours prior to triangulation 
used for skeleton 
and subsequent 
extraction and TIN 
gap filling, line 
program package has been developed for 
DTM, contour lines, skeleton, triangulation, distance transform, tessellation, interpolation, 
: Contours 
of fig. Al) 
supplying additional information about terrain 
relief at places of rapid slope change. In fact, 
several commercial DTM interpolation programs 
welcome breaklines and breakpoints, in addition to 
bulk data, in order to attain a better terrain 
relief reconstruction. Breaklines, structure lines 
  
and salient points can be conceived of as the 
skeleton of the terrain, sustaining regular 
surfaces. 
Differential Modelling 
Makarovic [9] proposed acquiring such skeleton 
information by photogrammetric selective sampling. 
The deficiencies of contours in difficult areas 
can be reduced and interpolation improved by 
combining contour data obtained from existing maps 
with skeleton data collected by photogrammetry. At 
the same time, supplementary samples at contour 
gaps can be incorporated. Tuladhar [13] reported 
improvement rates of DTM accuracy as great as 50%. 
The disadvantages of this approach are that 
photographs, photogrammetric equipment and DTM 
expertise are needed, and that the required manual 
operations are time-consuming. Sacrificing quality 
but gaining in convenience and costs will be an 
  
in 
ap 
ar 
lii 
sp 
ri 
an 
in! 
loc 
Ob; 
in! 
col 
auf 
re. 
wif 
the 
tr: 
fii 
(Se 
tog 
mal 
Fig 
Car 
exi 
str 
It 
tor 
str 
whe 
the 
wit 
inv 
lin 
app 
inf 
tra 
of 
ini 
thi 
rat 
For 
pla 
IBM 
pro 
ras 
Bor 
int 
inc 
dig 
pro