The Delaunay triangulation is the standard choice for building triangulated irregular networks (TINs) to represent terrain surfaces. However, the Delaunay triangulation is based only on the 2D coordinates of the data points, ignoring their elevation. This can affect the quality of the approximating surface. In fact, it has long been recognized that sometimes it may be beneficial to use other, non-Delaunay, criteria that take elevation into account to build TINs. Data-dependent triangulations were introduced decades ago to address this exact issue. However, data-dependent trianguations are rarely used in practice, mostly because the optimization of data-dependent criteria often results in triangulations with many slivers (i.e., thin and elongated triangles), which can cause several types of problems. More recently, in the field of computational geometry, higher order Delaunay triangulations (HODTs) were introduced, trying to tackle both issues at the same time—data-dependent criteria and good triangle shape—by combining data-dependent criteria with a relaxation of the Delaunay criterion. In this paper, we present the first extensive experimental study on the practical use of HODTs, as a tool to build data-dependent TINs. We present experiments with two USGS 30m digital elevation models that show that the use of HODTs can give significant improvements over the Delaunay triangulation for the criteria previously identified as most important for data-dependent triangulations, often with only a minor increase in running times. The triangulations produced have measure values comparable to those obtained with pure data-dependent approaches, without compromising the shape of the triangles, and can be computed much faster.
[1]
S. Rippa.
Long and thin triangles can be good for linear interpolation
,
1992
.
[2]
M. Garland,et al.
Fast Polygonal Approximation of Terrains and Height Fields
,
1998
.
[3]
C. Traas,et al.
Best data-dependent triangulations
,
2000
.
[4]
Hamid R. Arabnia,et al.
Comparison of existing triangulation methods for regularly and irregularly spaced height fields
,
2001,
Int. J. Geogr. Inf. Sci..
[5]
Gholamhossein Dastghaibyfard,et al.
Drainage reality in terrains with higher-order Delaunay triangulations
,
2008
.
[6]
Joachim Gudmundsson,et al.
Higher order Delaunay triangulations
,
2000,
Comput. Geom..
[7]
Jeffrey L. Brown,et al.
Vertex based data dependent triangulations
,
1991,
Comput. Aided Geom. Des..
[8]
J. Weisz,et al.
A refined “angle between normals” criterion for scattered data interpolation
,
2001
.
[9]
Maarten Löffler,et al.
Optimization for first order Delaunay triangulations
,
2007,
Comput. Geom..
[10]
Ivana Kolingerová,et al.
Multicriteria-optimized triangulations
,
2001,
The Visual Computer.
[11]
Larry L. Schumaker,et al.
Computing optimal triangulations using simulated annealing
,
1993,
Comput. Aided Geom. Des..
[12]
S. Rippa,et al.
Data Dependent Triangulations for Piecewise Linear Interpolation
,
1990
.
[13]
Bernd Hamann,et al.
On Simulated Annealing and the Construction of Linear Spline Approximations for Scattered Data
,
2001,
IEEE Trans. Vis. Comput. Graph..