You are on page 1of 13

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/261359062

Subdivision Surface Algorithm Based on a Novel Hexagonal Vertex-To-Vertex


Subdivision Scheme

Article · December 2005

CITATIONS READS

0 414

1 author:

Yasser Abd El-Latif


Ain Shams University
22 PUBLICATIONS   49 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Perfect matching View project

Meshes Compression View project

All content following this page was uploaded by Yasser Abd El-Latif on 07 October 2014.

The user has requested enhancement of the downloaded file.


Subdivision Surface Algorithm Based on A Novel
Hexagonal Vertex-To-Vertex Subdivision Scheme
Fayed F. M. Ghaleb Yasser M. Abd El-Latif*

Ain Shams University


Faculty of Science
Math. & Comp. Sc. Dept.
Abbassia, Cairo, Egypt

environment. Validation of our approach is


Abstract given by discussion, analysis and an illustration
One problem in subdivision surfaces is the of some experimental results.
number of meshes grows quickly after every
subdivision step. The surface with huge number Keywords: Subdivision surfaces,
of meshes is difficult to manipulate. In this Computational Geometry, Subdivision scheme,
paper we propose a hexagonal scheme which Boundary control, Geometric modeling,
can be considered as a modification to Doo- Meshes, Computer Graphics and Animation.
Sabin scheme. We introduce recursive
subdivision rules for meshes with arbitrary 1. Introduction
topology, optimizing the surface continuity
given a minimal support area. These rules have Subdivision surfaces are powerful tools for
simplicity comparable to the [25] scheme. Doo- graphical modeling and animation because of
Sabin subdivision scheme defined three types of their scalability, numerical stability, simplicity
new faces created during the construction of the in coding, fast generation of high quality
new mesh namely: face-face, an edge-face, and surfaces and especially, their ability to represent
a vertex-face. Alternatively, our proposed free-form surfaces of arbitrary topology [1], [2].
vertex-to-vertex subdivision scheme produces Subdivision algorithms are also attractive
the following three types: face-face, an edge- because they are conceptually simple and can be
face, and a vertex-vertex. Consequently, easily modified to create surface features
increasing the smoothness of the surfaces and without making major changes to the algorithm.
decreasing the running time. Surfaces Nowadays the subdivision algorithms are
smoothness and complexity time of our commonly used in films to model the computer
proposed algorithm are analyzed and compared graphics scenes and characters [3] and their
with existing methods. The original contribution study is still gaining interest due to their close
and advantages of our proposed technique over relation to wavelet and multi-resolution analysis
existing methods are that, the proposed [30], [31], [32], [33], [35]. A number of
technique produces good- quality surface commercial systems use subdivision as a surface
approximations without too many faces during representation like Pixar`s Render-man,
the construction process, simple to implement, Nichimen`s Mirai, Micropace` Lightwave 3D,
reliable in which it can be applicable to any etc. Surface subdivision is a technique for
surface mesh, optimizing the surface continuity, generating smooth surfaces that has been
has certain desirable properties, achieves introduced quite some time ago by Catmull-
numerical stability for the surfaces and its Clark [4] and Doo-Sabin [5]. The Basic idea of
computations are fast enough for an interactive the subdivision surfaces [1], [2] is to calculate
_______________________________________ smoother surface from the current surface mesh.
*Yassermal2002@Yahoo.com In the beginning there is an initial control point
mesh, which is refined by the sequence of strips, which converges to a specific B-Spline.
successive calculation. The initial mesh can be Sederberg et al. [9] add support for sharp edges
arbitrary and it can contain any number of to the Catmull-Clark and Doo-Sabin schemes by
vertices and faces. Sequence of refinements is introducing non-uniform knot vectors. Sharp
based on set of rules. This kind of rule set is edges are created in Doo-Sabin surfaces by
called as subdivision scheme. A number of setting knot spaces along certain control edges
subdivision schemes have been propose so far to zero. N. Dyne et al. [10] proposed a butterfly
and may be sorted into two categories. subdivision scheme, it is an interpolating
Approximation techniques move all the vertices scheme that extended by Zorin et al. [11] to
of the mesh, whereas interpolation techniques guarantee the C1 continuity for arbitrary meshes.
keep the vertices of the control mesh Subdivision schemes, which operate on
unchanged. Approximation techniques have triangular meshes were proposed and discussed
been most extensively studied, as they tend to by Farin [12] and Loop [13]. Hoppe et al.
produce better shapes and are more flexible. In modified the subdivision rules to add borders to
contrast, interpolation techniques preserve the Loop surfaces [14]. They also provide support
vertices of the control mesh, but the fairness of for creases, by uncoupling the crease curve from
the surface is more difficult to control. The type the surface from both sides [14]. Although the
of subdivision scheme proposed in this paper is Loop scheme [13] has been tested, it can be
an approximation. adapted to other schemes. Mandal et al. [15]
have proposed a scheme for dynamic
The paper is organized as follows. Section 2, manipulation of the limit surface created with
reviews some of the prior researches in the modified Butterfly scheme [10], [11] using
subdivision surfaces and its challenging. Section physically based force controls. The control is
3, gives a brief summary of the Doo-Sabin provided by tracking vertices at various levels
subdivision scheme and its subdivision of the subdivision. Volino [16] uses blended
problems. Section 4, introduces our proposed spheres to perform a N-adic tessellation of
algorithm for surface subdivision. Some triangles. Points and normals from the initial
experimental results with discussions are given mesh are directly involved in its construction.
in sections 5. Analysis of our proposed Vlachos [17] uses a N-dic tessellation to
algorithm compared with existing methods is perform his PN Triangles (or N-patches). Later,
discussed in sections 6. Finally, in section 7, we a new class of interpolating N-adic subdivision
conclude this paper with some direction for algorithm has been proposed in [18]. Some
future work. other methods deal with the deformation of the
surface by modifying the normals at the vertices
2. Prior Research of the control mesh. For instance, Halstead et al.
[19] has improved the Catmull-Clark [4] scheme
Subdivision surfaces are rapidly gaining in order to directly generate the limit surface.
popularity in computer graphics and other Conditions for interpolating normals are given.
applied areas. A number of subdivision schemes Unfortunately, using normals to control the limit
have been proposed since Catmull-Clark [4] and surface dramatically increases the resolution
Doo-Sabin [5] introduced subdivision surfaces cost. Biermann and Levin provide the Catmull-
in 1978. Nasri [6] extended Doo-Sabin surface Clark and Loop schemes with flexible support
to interpolate vertices of original polyhedron for borders and normal control [20], [21]. They
and B-spline curves on the subdivision surface. also observed that in certain situations it is
Nasri [7], [8] further describes methods to create interesting to employ different rules for convex
creases and borders for the dual Doo-Sabin and concave borders. Levin [22] further
scheme. He augments the object with polygonal
2
extended the boundary to allow arbitrary
parametric curves for interpolating a network of
curves. Velho and Zorin add support for smooth
borders to their 4-8-subdivision scheme [23].
Kobbelt scheme [24] is basically like 4-8-
subdivision scheme [23] but the basic tiling is 3-
6. Recently schemes that operate on the dual of
triangular meshes – namely hexagonal meshes -
were investigated by Claes et al [25] and Beets Figure 1 Chaikin`s Algorithm
et al. [26]. Another adaptive subdivision
schemes have been proposed in Muller [27], Doo-Sabin adapted the refinement techniques
[28], [29]. Compared with the previous for the quadratic uniform B-spline surface and
mentioned work, this paper proposed a new developed a new surface definition procedure
subdivision scheme such a hexagonal scheme. based upon refinement. In Doo-Sabin method,
We introduce recursive subdivision rules for surfaces are generated from polyhedral
meshes with arbitrary topology, optimizing the networks by successively cutting the corners
surface continuity given a minimal support area. and edges of the polyhedron. Its procedure is
These rules have a simplicity comparable to the based on the observation that the points are
Doo-Sabin scheme. The advantages of our simply the average of four particular points
proposed technique are that, it produces good- taken in a polygon - the vertex for which the
quality surface approximations without too new point is being defined, the two edge points
many faces during the construction process, (the midpoints of the edges that are adjacent to
simple to implement, reliable in which it can be this vertex in the polygon), and the face point
applicable to any surface mesh, optimizing the (the centroid of the polygon).
surface continuity, has certain desirable
properties, achieves numerical stability for the
surfaces and its computations are fast enough
for an interactive environment.

3. Doo-Sabin Subdivision Surfaces and


Problems
The fundamental idea of Doo-Sabin subdivision
surfaces goes back to Chaikins algorithm [34],
which generates a quadratic B-spline curve from Figure 2 Doo-Sabin Subdivision Process
a polygon by successively cutting its corners.
Each subdivision generates two new points on Figure 2 illustrates such procedure, where the
each polygon leg at (1/4,3/4). For the polygon polygon face P 0P1P 2P3 is generated from the
shown in Figure 1 with n vertices (Vi)0<i<n+1 on polygon face P0 P1 P2 P3 as follows:
polygon leg (Vi,Vi+1), two new points are given
by the following:
V i   3 / 4 V i  1 / 4 V i  1 , V i  1  1 / 4 V i  3 / 4 V i  1 .
Then the new polygon is linked by the new
generated points (V i , V i  1 ) corresponding to
(Vi)0<i<n+1. By repeating the subdivision process,
we can get a quadratic B-spline curve.

3
through a set of rules called subdivision scheme.
The Doo-Sabin subdivision scheme defined
three types of new faces created during the
construction of the new mesh namely: face-face,
an edge-face, and a vertex-face. Figure 3
illustrates these new faces, which are created as
described below:

 Face-face (F-face): For each n-sided face, a


new small n-sided face is generated inside
that face. This new face is called F-face.
 Edge-face (E-face): For each edge,
connecting the new vertices adjacent to that
edge produces a four-sided face. This new
face is called E-face.
 Vertex-face (V-face): For each vertex,
In general, the equations to compute the new connecting the new vertices adjacent to a
points from the old ones are linear in the points vertex produces a new face. This new face is
and hence may be expressed in matrix form. For called V-face.
any regular face ABCD, we can obtain a new
face A  B C D  from the following expression:

A  B C D   S ( ABCD ),

where ,

9 3 1 3
 
3 9 3 1
S  1 / 16  
1 3 9 3
 
3 1 3 9
is the Doo-Sabin subdivision matrix for a
quadrilateral face. In triangle mesh, for any Figure 3 Three Types of Faces in Doo-Sabin surfaces.
regular face ABC, we can obtain a new face
A  B C " from the following expression: In the Doo-Sabin subdivision scheme and after
several subdivision steps, the generated
A  B C   S ( ABC ), subdivision surface will be smooth enough to
where , represent a fine shape. But even after few
subdivision steps, the number of generated
14 5 5 
meshes will become huge consequently difficult
 
S  1 / 24 5 14 5 to manipulate and the number of faces grows
 
 5 5 14  exponentially. Due to memory and time
restrictions, the number of subdivision steps that
is the Doo-Sabin subdivision matrix for a
can be performed is relatively small. The reason
triangle face.
for this problem is due to the three types (F-
face, E-face, and V-face) produced by the Doo-
The idea of the subdivision is to calculate a
Sabin subdivision scheme in which each vertex,
smoother surface from the current surface mesh
edge and n-sided face produces a new face
4
during the recursive construction process. For of the corner points of the cube. Let the points
instance, in the Doo-Sabin subdivision surfaces, P1, P2, P3 be the new face points corresponding
the number of meshes after one refinement step to vertex V created according to the Doo-Sabin
is about four times that of the original meshes. subdivision scheme. In our proposed
Thus, it is difficult to deal with these data, subdivision scheme, we calculate a new vertex
consequently adaptive techniques are still V` corresponding to vertex V instead of the face
necessary to manage the exponential growth of P1 P2 P3 by averaging the four points P1, P2, P3
the number of new polygons created during the and V as shown in Figure 4. This process is
subdivision process. This paper proposes such repeated for every construction process.
an adaptive technique to produce a hexagonal
scheme which we call vertex-to-vertex
subdivision scheme. The advantages of this
scheme are that, it generates soft meshes with
less polygon than Doo-Sabin method, generates
a hexagonal scheme without need to suitable
methods to convert a triangular to a hexagonal
as in [25], produces good-quality surface Figure 4 Vertex-to-Vertex illustrations with our
proposed algorithm.
approximations without too many faces during
the construction process, has certain desirable The pseudo-code of our proposed algorithm is
properties, efficient in time complexity, simple given as follows:
to implement, and achieves numerical stability
for the surfaces. Procedure Subdivision

4. Our Proposed Vertex-to-Vertex Stage 1: For each face f do


Subdivision Scheme - Calculates the set of new points
in this face f by the Doo-Sabin
The problem with the Doo-Sabin subdivision process;
scheme is that the subdivision process produces - Connects these new points in
a sequence of meshes with increasing number of order to construct a new face;
faces. These faces grow quickly after every - Add the necessary new edges;
subdivision step, consequently difficult to End;
manipulate. In this section an adaptive process Stage 2: For each vertex v do
for reducing the number of faces after every - Calculate a new point
subdivision step with the Doo-Sabin subdivision corresponding to the vertex v
scheme is presented. The Doo-Sabin subdivision by averaging all of the new
scheme defined three types of new faces created points that have been generated
during the construction of the new mesh for the faces that are adjacent
namely: face-face, an edge-face, and a vertex- to this vertex v including itself;
face. Alternatively, our proposed vertex-to- End;
vertex subdivision scheme produces the Stage 3: For each edge e do
following three types: face-face, an edge-face, If the two adjacent faces for e
and a vertex-vertex. The objective of the have not the same normal then
proposed approach is to optimize the surface - Construct a new face by
continuity and to manage the exponential connects the new points that
growth of the number of new faces created by have been generated for the
the subdivision process. Let us consider V one

5
faces and vertices that are L
adjacent to this edge e;
Otherwise,
- Construct a new face by
connects the new points that
have been generated for the 0
faces only that are adjacent to
this edge e;
End if;
- Add the necessary new edges;
End; 1
The polygons generated through this refinement
step become the input set of polygons for the
next iteration step.

5. Experimental Results and Discussion


Based on our proposed algorithm, several 2
experimental results with animation have been
performed over an arbitrary initial control mesh.
Figure 5(a), illustrates a pentagon and it’s the
Doo-Sabin surfaces in different subdivision
steps. Figure 5(b) shows an animation sequence
of the same pentagon using our proposed
subdivision algorithm. Figures 6 demonstrate 3
other smooth subdivision process with the
second level from an arbitrary initial control
mesh using our proposed subdivision algorithm.
From the experimental results one can see that
in compared with the Doo-Sabin Subdivision
algorithm, our proposed subdivision algorithm
posses the following advantages. First, produces 4
good-quality surface approximations without
too many faces during the construction process.
Second, optimizing the surface continuity and
has certain desirable properties. Third, achieves
numerical stability for the surfaces. Finally, the
proposed technique is efficient in time
complexity and simple to implement, 5
consequently, fast enough for an interactive
environment. Such advantages are proved in
terms of the analysis discussed in the next
(a) (b)
section.
Figure 5 An illustration of the subdivision process of
the pentagon using (a) Doo-Sabin algorithm and (b)
our proposed subdivision algorithm.

6
algorithm is to construct the connectivity of the
output mesh correctly, and one would like to
accomplish this task efficiently. As discussed in
section 3 and shown in Figure 3, the Doo-Sabin
subdivision scheme defined three types of new
faces created during the construction of the new
mesh namely: (F-face, E-face and V-face). After
few subdivision steps, the number of generated
meshes will become huge and the number of
faces grows exponentially, consequently
difficult to manipulate. Compared with the Doo-
Sabin subdivision scheme, our proposed
algorithm reducing the number of faces after
every subdivision step and only creates two face
types F-face and E-face. The V-face type is lead
to another vertex and not face, consequently,
generating a hexagonal scheme. This scheme
will increase the smoothness of the surfaces and
decrease the running time comparing with Doo-
Sabin scheme. Also this scheme prevents the
coarse area that occurs with Doo-Sabin scheme
see Figure 5 with level 5.

Figure 7. A subdivision of a regular mesh with a factor


of three and four.

It turns out there exist many ways in which a


hexagonal grid can be recursively refined. The
first two non-trivial of the possible subdivisions
Figure 6 Subdivision surfaces with an arbitrary
multiply the number of hexagons by
control mesh respectively three and four. A scaling factor of
three is investigated in a paper by Johan Claes,
6. Analysis Koen Beets, and Frank Van [25]. A scaled-
down version of the input hexagon is dropped in
The subdivision process produces a sequence of its center, after which the new vertices are
polyhedron with increasing numbers of faces reconnected to the nearest new vertices,
and vertices. Subdivision surfaces are the limit replacing each old vertex by a new hexagon (as
of this sequence. This section analyzes the in the left image of Figure 7). This can be
surfaces smoothness of our proposed algorithm viewed as an corner-cutting operation. This
compared with the [25], and Doo-Sabin setup introduces one type of new vertex as in
subdivision algorithms. The key of the the Doo-Sabin scheme and every subdivision

7
step removes the vertices of the previous steps.
However, contrary to Doo-Sabin’s scheme, our
hexagonal scheme needs to keep the existing
vertices. So two types of rules are needed: one
for the new vertices, and possibly one for
relaxing the position of the old ones. Such a
scheme with a multiplication factor of four
would be the dual of triangular schemes that
split triangles into four, such as Charles Loop’s
Figure 9. Filter masks for V-vertex (vertex mask) in
scheme [13]. As this scheme for hexagonal case triangles mesh with (a) Interior vertex and (b)
subdivision does not have a fixed name yet, we Exterior vertex.
propose to call it the hexagon vertex-to-vertex
scheme. The smoothing operator can be implemented
very efficiently using repeated convolution by a
6.1 Smoothness sequence of averaging operations. In a first pass,
F-vertices are calculated as Doo-Sabin method.
After the application of the refinement rule to In a second pass, V-vertices are updated as the
the complex mesh M = (V, E, F), the set of average of their current value and the average of
newly inserted vertices V` of M`, can be the values of new vertices in their 1-
naturally divided into two classes: new vertices neighborhood. It is easy to verify that this
corresponding to each face which we will call procedure corresponds to the stencils in Figures
F-vertices, and new vertices corresponding to 8, 9.
each vertex which we call V-vertices. The
smoothing operator is a convolution filter. It The masks for quadrilateral or triangle face can
uses a different smoothing rule for each class of be calculated by using Doo-Sabin matrix as
vertices. describe in section 3; vertex mask coefficients
depend on k, the number of incident block faces.
The stencil for F-vertices is depicted as used in In the case shown in the figures 8(a), 9(a), k = 7.
Doo-Sabin method (face mask). The stencil for
V-vertices is depicted in Figures 8, 9 (vertex However, it is often necessary to model surfaces
mask). Observe that the filter kernel extends with boundary, which may contain sharp
beyond the first neighbors of the vertex. It features as well. Thus, it is of practical
corresponds to a smoothing filter based on 1- importance to extend our subdivision scheme to
neighbors before the introduction of new support surfaces with smooth boundaries and
vertices. creases. Furthermore, it is often useful to have
surfaces with piecewise smooth boundary. Here
we present only the rules for smooth boundaries.

Figures 8(b), 9(b) illustrate the V-vertex belongs


to the boundary quadrilateral and triangle mesh
respectively, vertex mask coefficients depend on
k, the number of incident block faces. In the
case shown in the figures, k = 4 for quadrilateral
Figure 8. Filter masks for V-vertex (vertex mask) in and triangle meshes. As we see that the same
case quadrilateral mesh with (a) Interior vertex and rules have to be applied in the neighborhood of
(b) Exterior vertex. corner vertices on the boundaries and creases,

8
except the adjacent boundary vertices for this  13 4 1 1 1 4 
vertex.  
 4 13 4 1 1 1 
 1 4 13 4 1 1 
A geometric interpretation of our scheme is 1
M   
6
available which we now employ to analyze the 24  1 1 4 13 4 1 
rule for quadrilateral and 6-gon mesh. Let P 1 ,  
1 1 1 4 13 4
 
P 2 , P 3 and P 4 be the four vertices of a regular  
 4 1 1 1 4 13 
square face as shown in Figure 10(a). A new set
of vertices q 1 , q 2 , q 3 , and q 4 , is generated by is the subdivision matrix for a triangle face.
application of the simple linear expression This matrix has the following eigenvalues:
 q 1 , q 2 , q 3 , q 4   M 4  P1 , P2 , P3 , P4  where
T T

 0  1

9 3 5

3 1
  1 ,  2 
8
1 3 9 3 1
M  3
4  3  3,4 
16 1 3 9
  8
 
3 1 3 9
1
 5 
4
is the subdivision matrix for a regular face of 4-
sided. In the literature [5], [36], [2], the following is
This matrix has the following eigenvalues: reported about the nature of the eigenvalues of
the subdivision matrix M 6 :
 0  1
1 o The largest eigenvalue should be 1. This is
 1 ,  2 
2 necessary for the scheme to be affine
1 invariant.
 3 
4
o The second and third largest eigenvalues
In similar way, we can apply this refinement (sorted by their absolute values) should
rule for 6-Gon faces. Let P 1 , P 2 , P 3 , P 4 , P 5 and be equal, and they should be strictly
larger than the next eigenvalues for the
P 6 be the vertices of a 6-Gon face as shown in
scheme to be C1 .
Figure 10(b). A new set of vertices q 1 , q 2 , q 3 ,
q 4 , q 5 and q 6 , is generated by application of o The fourth and fifth eigenvalues should be
the simple linear expression equal, and this indicates that there will be no
tangent vector at the central point.
 q 1 , q 2 , q 3 , q 4 , q 5 , q 6   M 6  P1 , P2 , P3 , P4 , P5 , P6 
T T

o These eigenvalues are related to the out-of-


where the-plane behavior of the configuration.

9
that can be generated in Doo-Sabin and manage
the exponential growth of the number of new
polygons created during the subdivision process.
The original contribution and advantages of the
proposed technique compared with previous
techniques are that, the proposed technique
produces good-quality surface approximations
without too many faces during the construction
process, simple to implement, reliable in which
(a) it can be applicable to any surface mesh,
optimizing the surface continuity, has certain
desirable properties, achieves numerical
stability for the surfaces and its computations
are fast enough for an interactive environment.
We are currently working on a rendering system
for both interactive and realistic rendering that
only generates these parts of the geometry that
are visible and disposes of the mesh parts that
(b) have already been rendered.
Figure 10. Filter masks for face mask in case (a)
quadrilateral mesh or (b) 6-Gon mesh. References
7. Conclusions and Future Work [1] Subdivision for modeling and animation.
SIGGRAPH 2000 Course notes.
Subdivision surfaces are efficient tools for fast
[2] M. Sabin, “Subdivision: Tutorial notes”,
generation of high quality surfaces and have
Tutorial of IEEE Shape Modelling and
been in the highlights of the computer graphics
Applications International Conference
community for the past few years. Nowadays,
(SMI`2001), May 2001.
the subdivision algorithms are commonly used
[3] T. DeRose, M. Kass, and T. Truong,
in films to model the computer graphics scenes
“Subdivision surfaces in character
and characters. The problem with the
animation”, SIGGRAPH`1998, pp. 85-94,
subdivision algorithms are that after few
1998.
subdivision steps, a number of generated
[4] Ed Catmull and J. Clark, “Recursively
meshes will become huge and the number of
generated B-Spline surfaces on arbitrary
faces grows exponentially, consequently
topological meshes”, Computer Aided
difficult to manipulate. Due to memory and time
Design, Vol. (10), No. 6, pp.350-355, 1978.
restrictions, the number of subdivision steps that
[5] D. Doo and M. Sabin, “Analysis of the
can be performed is relatively small. For
behaviour of recursive division surfaces
instance, in the Doo-Sabin subdivision surfaces,
near extraordinary points”, Computer Aided
the number of meshes after one refinement step
Design, Vol (10), No. 6, pp.356-360, 1978.
is about four times that of the original meshes.
[6] A. Nasri, “Polyhedral subdivision methods
This paper proposes a new subdivision
for free-form surfaces”, ACM Trans. On
technique called vertex-to-vertex subdivision
Graphics, Vol (6), No. 1, pp.29-73, 1987.
scheme as a key solution for such a challenge.
[7] A. Nasri, “Interpolating meshes of
The objective of the proposed technique is to
boundary intersecting curves by subdivision
reduce the huge number of meshes generated in
the Doo-Sabin surfaces, prevent the coarse area

10
surfaces”, The Visual Computer, Vol (16), Shape Modeling and Applications
No. 1, pp.3-14, 2000. International Conference, pp. 21-28, 2002.
[8] A. Nasri, “Constructing polygonal [19] M. Halstead, M. Kass and T. DeRose, “Fair
complexes with shape handles for curve Interpolation using Catmul-Clark surfaces”,
interpolation by subdivision surfaces”, SIGGRAPH`1993 Proceeding, pp. 35-44,
Computer Aided Design, Vol (33), 1993.
pp.753-765, 2001. [20] H. Biermann, A Levin and D Zorin,
[9] T. Sederberg, J. Zheng, J. Sewell and M. “Piecewise smooth subdivision surfaces
Sabin. “Non-Uniform recursive subdivision with normal control”, SIGGRAPH`2000
surfaces ”, SIGGRAPH`1998, pp. 387-394, Conf. Proc., pp. 113-120, 2000.
1998. [21] A. Levin, “Combined subdivision schemes
[10] N. Dyne, D. Levin and J. Gregory, “A for the design of surfaces satisfying
Butterfly subdivision scheme for surface boundary conditions”, Computer Aided
interpolation with tension control”, ACM Geometric Design, Vol (16), No. 5, pp.345-
Trans. On Graphics, Vol (9), No. 2, pp.160- 354, 1999.
169, 1990. [22] A. Levin, “Interpolating nets of curves by
[11] D. Zorin, P. Schroder, and et. Al., smooth subdivision surfaces”,
“Interpolating subdivision for meshes with SIGGRAPH`1999 Proceeding, pp. 57-64,
arbitrary topology”, SIGGRAPH`1996, pp. 1999.
189-192, 1996. [23] L. Velho and D. Zorin, “4-8 Subdivision”,
[12] G. Farin “Designing C1 surfaces consisting In Computer Aided Geometric Design, Vol
of triangular cubic patches”, Computer (16), No. 5, pp.397-427, 2001.
Aided Design, Vol (14), pp.253-256, 1982. [24] L. Kobbelt, “Sqrt(3) Subdivision”,
[13] C. Loop, “Smooth subdivision surfaces SIGGRAPH`2000 Conf. Proceeding, pp.
based on triangles”, Univ. of Utah, Math. 103-112, 2000.
Dept., Master Thesis, 1987. [25] J. Claes, K. Beets, and V. Frank, “A
[14] H. Hoppe, T. DeRose, et. Al., “Piecewise Corner-cutting Scheme for hexagonal
smooth surface reconstruction”, Subdivision Surfaces”, Proc. of IEEE Shape
SIGGRAPH`1994, pp. 295-302, 1994. Modeling and Applications International
[15] C. Mandal, H. Qin, and C. Vermuri, Conference, pp. 13-20, 2002.
“Dynamic modeling of Butterfly [26] K. Beets, J. Claes, and V. Frank, “Borders,
Subdivision surfaces”, IEEE Trans. semi-Sharp Edges and Adaptivity for
Visualisation and Computer Graphics, Vol Hexagonal subdivision surface schemes ”,
(19), No. 3, pp. 427-436, 2000. Advances in Modeling, Animation and
[16] P. Volino and N Magenant Thalmann, “The Rendering, Springer-Verlag, pp. 151-166,
Spherigon: a simple polygon patch for 2002.
smoothing quickly your polygonal meshes”, [27] H. Mueller and R. Jaeschke, “Adaptive
Proc. of IEEE Computer Animation`98, pp. subdivision curves and surfaces”, Proc.
72-79, 1998. IEEE Computer Graphics International
[17] A. Vlachos, J. Peters, C. Boyed and J. Conference, pp. 48-58, 1998.
Michell, “Curved PN Triangles”, ACM [28] L. Kobbelt, “Interpolatory subdivision on
Symposium on Interactive 3D Graphics, pp. open Quadrilateral nets with arbitrary
159-166, 2001. topology”, Computer Graphics Forum,
[18] G. Salomon, A. Leclercq, S. Akkouche and EuroGraphics`1996, pp. 409-420, 1996.
E. Gallin, “Normal Control using N-adic [29] Z. Xu, and K. Kondo, “Adaptive
Subdivision Schemes”, Proc. of IEEE Refinements in Subdivision Surfaces”,

11
Computer Graphics Forum,
EuroGraphics`1999, pp. 239-242, 1999.
[30] D. Zorin, P. Schroder and W. Sweldens,
“Interactive Multi-resolution Mesh
Editing”, SIGGRAPH`1997 Proceeding, pp.
159-168, 1997.
[31] I. Guskov, W. Sweldens, and P. Schroder,
“Multi-resolution signal processing for
meshes” SIGGRAPH`99, pp. 325-334,
1999.
[32] L. Kobbelt et al., “Interactive multi-
resolution modeling on arbitrary meshes”,
SIGGRAPH`98, pp. 105-115, 1998.
[33] B. Jawerth and W. Sweldens , “An
overview of wavelet based multi-resolution
analysis”, SIAM Review, Vol. 36, No. 3,
pp. 377-412, 1994.
[34] G. M. Chaikin, “An Algorithm fo High-
Speed Curve Generation”, Computer
Graphics and Image Processing, 3, pp. 346-
349, 1974.
[35] U. Reif, “A Unified Approach to
Subdivision Schemes near Extraordinary
Vertices,” in Computer Aided Geometrical
Design, 12(2), pp.153-174, 1995.

12

View publication stats

You might also like