Comparative Study of Mesh Simplification

Algorithms

Abderazzak Taime, Abderrahim Saaidi and Khalid Satori

Abstract Many applications in the field of computer graphics are becoming more
complex and require more accurate simplification of the surface meshes. This need
is due to reasons of rendering speed, the capacity the backup and the transmission
speed 3D models over networks. We presented four basic methods for simplifying
meshes that are proposed in recent years. The result obtained by the implementation
of these methods will be the subject of a comparative study. This study aims to
evaluate these methods in terms of preserving the topology and speed.

Keywords Computer graphics
Simplification
Surface meshes

1 Introduction

The use of 3D models, represented as mesh, is ever increasing in many applications

and are also used in very different fields, such as video games, animation, archeology,
medical imaging, scientific visualization, augmented reality, and computer aided
design. These models are finely detailed to satisfy an expectation of increasingly

A. Taime (&)
A. Saaidi
K. Satori

LIIAN, Department of Mathematics and Computer Science Faculty of Sciences,
Dhar-Mahraz Sidi Mohamed Ben Abdellah University, B.P 1796 Atlas- Fez, Morocco
e-mail: [email protected]
K. Satori
e-mail: [email protected]
A. Saaidi
LSI, Department of Mathematics, Physics and Computer Science Polydisciplinary
Faculty of Taza, Dhar-Mahraz Sidi Mohamed Ben Abdellah University,
B.P. 1223 Taza, Morocco
e-mail: [email protected]

© Springer International Publishing Switzerland 2016 287

A. El Oualkadi et al. (eds.), Proceedings of the Mediterranean Conference
on Information & Communication Technologies 2015, Lecture Notes
in Electrical Engineering 380, DOI 10.1007/978-3-319-30301-7_30
288 A. Taime et al.

strong realism in computer graphics. However, the increased complexity of the data
still surpasses the improvements in hardware performance, which requires the sim-
plification of meshes while preserving the topology. In this context, several
approaches and strategies are adopted for controlled simplification of meshes:
• The first Involves the removal of geometric entities, typically the vertices [1] or
triangles [2] and the remesh the holes thus formed.
• Some strategies have tried to agglomerate geometrically the nearby vertices to
replace an representative [3–8].
• Another type of method is to merge coplanar facets [9–11].
• Other approaches have tried optimizing an energy function [12, 13].
• Wavelets [14, 15], in a single structure, provide a mathematical framework to be
several representations more or less detailed of the same object.
Another type of method called remeshing consists of randomly positioning the
points on the surface and move them with attraction/repulsion forces by forcing
them to stay on the surface [16].
We can still classify these methods on the basis of local and global approaches:
in the first case, mesh modifications are operated upon a local optimization criterion
(e.g. simplification envelopes [17] and other approaches decimation). In the second,
a global optimization process is applied to the whole mesh. (e.g. energy opti-
mization approaches [12, 13], Multiresolution decimation [18, 19], and multires-
olution analysis [20]).
This work presents a brief introduction to mesh simplification methods. Its main
objective is to analyze and compare some methods and main approaches on the
basis of two criteria, a criterion for minimizing the error or difference between the
simplified mesh and the original mesh and a criterion of run times.

2 Error Evaluation

The approximation error is managed in many various manners by the different

simplification approaches, but there are two principal manners between them:
• Approaches that are based on the locally bounded errors, i.e. the approximation
accuracy is known in each surface unit (e.g. methods of mesh decimation
[1, 21]).
• Approaches which only support globally bounded approximation errors, i.e. the
approximation accuracy is known around each surface entity (e.g. the simpli-
fication envelopes method [17], Agglomeration of the vertices [3, 4, 8] and
energy optimization approaches [12, 13]).
Comparative Study of Mesh Simplification Algorithms 289

3 Simplification Approaches Reviewed

To quickly view a scene and allow it to move, the mesh is simplified by reducing
the number of triangles. The strategy for simplifying a mesh differ from one
approach to another, On the other hand, the majority of them degrade the mesh
approximation precision to reduce the number of vertices/triangles. Here is an
overview of approaches and the most commonly used methods for simplifying
meshes we presented in this work:
• Vertex decimation: This algorithm iteratively removes a vertex of the mesh and
all the adjacent faces. In each step of decimation process, all vertices are
evaluated according to their importance. The least important vertex is selected
for removal and all the facets adjacent to that vertex are removed from the model
and the resulting hole is triangulated (see [1] for more details). Since the tri-
angulation requires a projection of the local surface onto a plane, these algo-
rithms are generally limited to manifold surfaces. Vertex decimation methods
preserve the mesh topology as well as a subset of original vertices.
• Energy function optimization: the two methods most usually used within this
framework are those proposed by Hoppe [12, 13]. The principle of these
methods consists in assigning an energy function to three characteristics of the
mesh: the number of nodes, the approximation error and the regularity of the
mesh (determined from the length of the edges). These methods seek to mini-
mize the energy in order to solve the mesh optimization problem. This approach
can produce much better simplification results, but it increases a huge amount of
computational cost.
• Agglomeration of the vertices: the methods based on this principle are distin-
guished by how to choose the vertices to be agglomerated. There are essentially
two ways to do it. The first uses a three-dimensional grid, possibly adaptive, to
replace all of the vertices belonging to a cell of the grid with a single vertex. [3]
The second approach uses a metric for determining the nodes to be merged [4]
Fig. 1.

Contract

Before After

Fig. 1 Edge contraction. The highlighted edge is contracted into a single point. The shaded
triangles become degenerate and are removed during the contraction
290 A. Taime et al.

In this work we present a comparative study of four methods [1, 3, 4, 12]

representative of the approaches mentioned above. These methods are most com-
monly used in the literature and a source of inspiration for many works published in
recent years.

4 Evaluation of the Simplified Meshes

In order to give a theoretical assessment but also objective of the methods exam-
ined, we adopted an empirical approach based on four implementations in which
the approximation error introduced in the simplification process is calculated. For
this purpose, we have based on a metric of the approximation error used by Garland
[4], capable of measuring the average squared distance between the approximation
and the original model. The error approximation Ei = E (M, E) is defined as:
!
1 X
Ei ¼ d ð#; Mi Þ þ d ð#; Mn Þ
2 2
ð1Þ
jXn j þ jXi j #2Xn

where Xn and Xi are sets of points sampled on the models Mn and Mi respectively.
The distance dð0; MÞ ¼ minp2M jj#  pjj is the minimum distance from v to the
closest face of M.
Moreover, On the basis of the maximum error like metric associated to the
faces/vertices [21], in order to have a more precise idea of these methods.

5 Empirical Evaluation

To validate the effectiveness of these methods, we choose a set of data representing

a broad category (Fig. 2): the meshes obtained with standard CAD system—
SOCKET, available at http://www1.cs.columbia.edu/*cs4162/models/.
The four methods are executed on a computer with the following features: Core
2 Duo processor, 1.60 GHz, 2 GB RAM.
Table 1 shows the numerical results obtained by the implementation of the four
methods. The results relative to the evaluation of the approximation error are also
summarized in the Figs. 3 and 4 (see also Fig. 5): the Fig. 3 shows the maximal
error of the various methods and sizes. Similarly, Fig. 4: on the top graph, we plot
the error curve (EAVG) for different approaches and sizes. For good visualization,
we plot, in the lower graph, the error curve (EAVG) for three approaches and
Comparative Study of Mesh Simplification Algorithms 291

Fig. 2 A CAD triangulated mesh: SOCKET original—Meshing: 836 vertices

Table 1 Comparison of SOCKET (826 vertices, 1696 triangles, bounding box:

various simplification 4.8 × 5.7 × 2.8)
algorithms: SOCKET mesh,
NVert NTriang EMax EAvg Time
(errors are measured as
percentages of the datasets Mesh decimation [1]
bounding box diagonal; times 634 (75 %) 1302 0.00092 1.081e-0.5 218.05
are in milliseconds) 412 (50 %) 852 0.00532 5.091e-0.5 239.89
200 (25 %) 424 0.11553 0.00367 250.23
106 (12 %) 218 0.59810 0.01339 254.66
37 (4 %) 71 1.67380 0.04800 262.08
Vertex clustering [3]
634 (75 %) 1302 2.18201 0.03201 43.05
413 (50 %) 852 3.19371 0.16009 41.89
202 (25 %) 424 5.00101 0.54008 45.23
107 (12 %) 218 7.98261 1.80041 44.66
36 (4 %) N/A N/A N/A N/A
Mesh optimization [12]
635 (75 %) 1302 0.22921 0.00237 69,760
415 (50 %) 852 0.30975 0.00270 76,827
203 (25 %) 424 0.41219 0.00295 79,876
109 (12 %) 218 0.40520 0.00348 79,754
37 (4 %) 71 0.45591 0.00597 83,645
Quadric error metrics [4]
635 (75 %) 1302 0.00009 3.001e-0.5 298.05
413 (50 %) 852 0.00067 6.073e-0.5 327.79
200 (25 %) 424 0.00442 0.00022 341.13
106 (12 %) 218 0.03746 0.00173 344.50
37 (4 %) 71 0.09253 0.00571 351.18
292 A. Taime et al.

Fig. 3 The graphs show the Socket Mesh -- Max Error

performance of four 10 Decimation
approaches (E.MAX)
Vert. clustering
5
Mesh Optim.

0 Quadric Metrics
68. 204. 424. 848. 1272.
no. faces

Fig. 4 The graphs show the

performance of four
approaches (EAVG)

different sizes. The size of the simplified mesh is represented on the X axis and the
error on the Y axis.
In particular, the method of Vertex clustering [3] does not reach a high level of
simplification. Moreover, she recorded in general the worst result in terms of error,
but the maximum error curve showed a relative improvement; this depends on the
strategy adopted in the removal of the vertices. By cons, it is interesting to note that
this method produces the best results when speed is needed.
As expected, the good results in terms of average error are given by the Mesh
Optimization method and Quadric Error Metrics. However, the method of Mesh
decimation is less reliable in terms of accuracy. On the other hand, methods [3, 4,
12] which only support globally bounded approximation errors and which use a
global optimization process at mesh modifications produce better results when we
consider the maximal error.
It is interesting to note that the method of QEM and Mesh decimation are
produced the better results in terms of speed. In contrast, the method of Mesh
Optimization is the slowest.
Comparative Study of Mesh Simplification Algorithms 293

Mesh decimation ROSSIGNAC 93

Mesh optimization Quadric error metrics

Fig. 5 Simplified Socket meshes (*852 faces)

6 Conclusion

This work presented a brief overview of mesh simplification methods proposed in

recent years. We presented the different characteristics of the four basic methods
that are based on the simplification strategy. We also discussed and classified these
methods according to their ability to preserve topology and manage the errors. In
addition, the results of an empirical comparison have been presented. Four
implementations have been performed on a set of data. We compared the compu-
tational time and accuracy of simplified meshes.
From the point of view of accuracy, the results showed that approaches QEM
and Mesh Optimization gives the best results in terms of error.
Finally, all tested solutions have a common weakness: they are defined to work
on a single mesh. This is not the case of CAD models or virtual reality sessions
where we may need to simplify the complex objects or scenes with a
non-topological clean composition between components. New solutions are
required for these applications to provide increased generality and robustness.
294 A. Taime et al.

References

1. Schroeder, W.J., Zarge, J.A., Lorensen, W.E.: Decimation of triangle meshes. ACM siggraph
computer graphics, pp. 65–70. ACM, New York (1992)
2. Taubin, G., Guéziec, A., Horn, W., Lazarus, F.: Progressive forest split compression. In:
Proceedings of the 25th annual conference on Computer Graphics and Interactive Techniques.
ACM, New York, pp. 123–132 (1998)
3. Rossignac, J., Borrel, P.: Multi-resolution 3D approximations for rendering complex scenes,
pp. 455–465. Springer, Heidelberg (1993)
4. Garland, M., Heckbert, P.S.: Surface simplification using quadric error metrics. In:
Proceedings of the 24th annual Conference on Computer Graphics and Interactive
Techniques. ACM Press/Addison-Wesley Publishing Co. pp. 209–216 (1997)
5. Kanaya, T., Teshima, Y., Kobori, K.I., Nishio, K.: A topology-preserving polygonal
simplification using vertex clustering. In: Proceedings of the 3rd International Conference on
Computer Graphics and Interactive Techniques. ACM Australasia and South East Asia,
pp. 117–120 (2005)
6. Boubekeur, T., Alexa, M.: Mesh simplification by stochastic sampling and topological
clustering. Comput. Graph. 33(3), 241–249 (2009)
7. Li, Y., Zhu, Q.: A new mesh simplification algorithm based on quadric error metrics. In:
International Conference on Advanced Computer Theory and Engineering, ICACTE’08,
IEEE, pp. 528–532 (2008)
8. Li, G., Wang, W., Ding, G., Zou, Y., Wang, K.: The edge collapse algorithm based on the
batched iteration in mesh simplification. In: IEEE/ACIS 11th International Conference on
Computer and Information Science (ICIS), 20, IEEE, pp. 356–360 (2000)
9. Dehaemer, M.J., Zyda, M.J.: Simplification of objects rendered by polygonal approximations.
Comput. Graph. 15(2), 175–184 (1991)
10. Hinker, P., Hansen, C.: Geometric optimization. In: Proceedings of the 4th Conference on
Visualization’93. IEEE Computer Society, pp. 189–195 (1993)
11. Kalvin, A.D, Haddad, B., Noz, M.E.: Constructing topologically connected surfaces for the
comprehensive analysis of 3-D medical structures. In: Med. Imaging V. Image Process. Int.
Soc. Opt. Photonics, 247–258 (1991)
12. Hoppe, H., Derose, T., Duchamp, T., McDonald, J., Stuetzle, W.: Mesh optimization. In:
Proceedings of the 20th Annual Conference on Computer Graphics and Interactive
Techniques. ACM, New York, pp. 19–26 (1993)
13. Hoppe, H.: Progressive meshes. In: Proceedings of the 23rd Annual Conference on Computer
Graphics and Interactive Techniques. ACM, New York, pp. 99–108 (1996)
14. Eck, M., Derose, T., Duchamp, T., et al.: Multiresolution analysis of arbitrary meshes. In:
Proceedings of the 22nd Annual Conference on Computer Graphics and Interactive
Techniques. ACM, New York, pp. 173–182 (1995)
15. Wang, W., Zhang, Y.: Wavelets-based NURBS simplification and fairing. Comput. Methods
Appl. Mech. Eng. 199(5), 290–300 (2010)
16. Turk, G.: Re-tiling polygonal surfaces. ACM SIGGRAPH Comput. Graph. 26(2), 55–64
(1992)
17. Cohen, J., Varshney, A., Manocha, D., Turk, G., Weber, H., Agarwal, P., Wright, W.:
Simplification envelopes. In: Proceedings of the 23rd Annual Conference on Computer
Graphics and Interactive Techniques. ACM, New York, pp. 119–128 (1996)
18. Ciampalini, A., Cignoni, P., Montani, C., Scopigno, R.: Multiresolution decimation based on
global error. Visual Comput. 13(5), 228–246 (1997)
19. Mocanu, B., Tapu, R., Petrescu, T., Tapu, E.: An experimental evaluation of 3D mesh
decimation techniques. In: 10th International Symposium on Signals, Circuits and Systems
(ISSCS), IEEE, pp. 1–4 (2011)
Comparative Study of Mesh Simplification Algorithms 295

20. Cohen, A., Dyn, N., Hecht, F., Mirebeau, J.M.: Adaptive multiresolution analysis based on
anisotropic triangulations. Math. Comput. 81(278), 789–810 (2000)
21. Ciampalini, A., Cignoni, P., Montani, C., Scopigno, R.: Multiresolution decimation based on
global error. Visual Comput. 13(5), 228–246 (1997)