Conformal Equivalence
Conformal Equivalence
Conformal Equivalence
Boris Springborn
TU Berlin
Peter Schrder
Caltech
Ulrich Pinkall
TU Berlin
Abstract
We present a new algorithm for conformal mesh parameterization. It is based on a precise notion of discrete conformal equivalence for triangle meshes which mimics the notion of conformal
equivalence for smooth surfaces. The problem of finding a flat
mesh that is discretely conformally equivalent to a given mesh
can be solved efficiently by minimizing a convex energy function,
whose Hessian turns out to be the well known cot-Laplace operator. This method can also be used to map a surface mesh to a
parameter domain which is flat except for isolated cone singularities, and we show how these can be placed automatically in order
to reduce the distortion of the parameterization. We present the
salient features of the theory and elaborate the algorithms with
a number of examples.
Keywords: Discrete Differential Geometry; conformal parameterization; conformal equivalence; discrete Riemannian metric;
cone singularities; texture mapping
Introduction
In this paper we present a definition for discrete conformal equivalence of triangle meshes and apply it to the problem of conformal mesh parameterization. Our approach arises from two basic
premises.
First, the discrete setting should parallel the smooth setting.
There the problem of finding a conformal parameterization for a
smooth surface in space is equivalent to finding a flat metric on
the surface that is conformally equivalent to the metric induced
by the embedding. Hence, we cast the parameterization problem
as one of finding conformally equivalent flat metrics for a given
metric.
Second, we treat the parameterization problem for triangle
meshes in the spirit of Discrete Differential Geometry [Bobenko
and Suris 2005]: instead of viewing discretization as a means of
making the smooth problem amenable to numerical methods, we
seek to develop on the discrete level a geometric theory that is
as rich as the analogous theory for the smooth problem. The aim
is to discretize the whole theory, not just the equations.
Instead of asking for an approximation of the smooth problem,
we are thus guided by questions like: What corresponds to a
Riemannian metric and Gaussian curvature in an analogous theory for triangle meshes? When should two triangle meshes be
considered conformally equivalent? We answer the first question
in the obvious way: the edge lengths and the angle defects at
vertices. The answer that we give to the second question is also
straightforward and intuitive. The striking point is that once we
Conformal Equivalence
(1)
(2)
(3)
ci j := l im /l m j l jk /l ki .
(4)
(5)
2.1
ijk = 2 tan1
(l i j +l jk l ki )(l jk +l ki l i j )
(l ki +l i j l jk )(l jk +l ki +l i j )
(6)
ij,
jk ,
ki ) (ui +u j +uk ) + 1
b i ui (7)
E(u) =
f (
2
2
t i jk T
vi V
where
ij,
jk ,
ki ) =
f (
j
j
1
jk +
i j + L(
ijk
ki
kij
ki ) + L(
ijk ) + L(
kij ),
ki +
2
(8)
These methods require the solution of non-linear equations, using a convex variational principle, a specially tailored relaxation
procedure, or a flow.
This approach to discrete conformality is also not fully satisfactory.
In the case of Kharevych et al., the intersection angles determine
the curvature of the mesh. Consequently, they must change when
the mesh is flattened. For the other circle methods, the notion
of discrete conformality either depends on an arbitrary choice
of circles, or it is defined only if the edge-lengths satisfy some
further conditions, or both.
Curvature Flow Consider u as a function of time and evolving
(9)
so that grad E(u) = 0 iff u solves the problem. The Hessian of the
energy turns out to be one half of the well-known cot-Laplace
operator:
P
(Hess E u)i = 12 (u)i = 14 ei j 3vi w i j (ui u j ),
(10)
kij + cot
li j for interior edges and only one cot
with w i j = cot
term for boundary edges. (We adopt the sign convention for
the Laplace operator which renders it positive semi-definite. See
Appendix B and C for proofs of Eqs. (9) and (10).)
The Hessian is therefore positive semi-definite with only the constant functions in its null-space. This corresponds to the fact that
the energy E(u) is scale invariant, i.e., its value does not change
if the same number is added to all ui .
2.2
(12)
To flatten a mesh, Ben-Chen et al. proceed by solving the discretized Poisson equation with the cot-Laplace operator in the
original metric and Ki the angle defect at a vertex. This is precisely the first Newton step when minimizing our energy E(u).
(Their definition of length change induced by the ui is different
from Eq. (2) but this is immaterial to the present argument.) As
Ben-Chen et al. point out, this does not in general yield a flat
metric. They address this by solving a cot-Laplace layout problem
in the new metric to find flat vertex positions which approximate
the desired metric. In contrast we solve for a (numerically) flat
metric and require no further approximation to recover the actual
vertex positions.
In their method, target curvatures at the boundary as well as for
cone singularities are computed based on a diffusion problem and
its associated Greens functionsone each per boundary vertex
and interior cone vertex. Incidentally, this computational burden
could be removed by using our natural boundary conditions and
free cone angles (Sections 4.2 and 5.1).
2.2.1
Discussion
solves the problem of finding polyhedral realizations for complete hyperbolic surfaces. But this side of the story is beyond the
scope of the present paper.
Now that the theoretical foundations are laid down we proceed with a discussion of our basic conformal parameterization
method.
3.1
Convex Optimization
then
ijk = and
ki =
kij = 0.
u3
u2
u1 +u2 +u3 =0
12 =23 =31 =0
We now have a standard unconstrained convex optimization problem with explicit formul for the target function, its gradient,
and Hessian. The Hessian is positive semi-definite with only the
constant vector in its null-space, reflecting the fact that E(u) remains invariant if a constant vector is added to u. To find the
argmin of the extended energy we use the globally convergent
Newton-Steihaug trust region method [Steihaug 1983] as implemented in PETSc/TAO [Balay et al. 2007; Benson et al. 2007].
3.2
3.3
Layout
geneous coordinates. The figure above shows a comparison between linear interpolation (top) and projective interpolation (bottom) with some notable differences highlighted.
Boundary Conditions
Both as a practical matter and as a key to achieving functionality relevant in applications, boundary conditions of different
types can be employed. We review these in turn, using practical
examples to motivate them.
4.1
Figure 2: A sampling of ripped edges. In all cases the local
geometry was near singular and edge flipping (alternatively, edge
subdivision) remedied the trouble.
data flat to within k grad E(u )k2 < = 1014 . . . 1012 and yield
worst relative length errors bounded by 10 to 1000.
Arbitrary topology meshes and those with cone singularities are
reduced to the topological disk case. To resolve handles we compute a system of loops [Erickson and Whittlesey 2005]. Cutting
the loops gives a topological disk, possibly with cones in its interior. To resolve the cones, we trace a path from each cone back to
the base vertex of the system of loops using the Dijkstra tree from
the first phase. (Any boundaries, including the one due to the
system of loops, are treated as if each were a single cone vertex
from the point of view of spanning tree construction.) Cutting
this cone spanning tree yields a topological disk with no cones
in the interior and it may be laid out as above. We have made
no efforts to move the paths to inconspicuous locations, though
methods such as Seamster [Sheffer and Hart 2002] could be employed. Note that vertices along the cut path will have multiple
positions in the layout.
3.4
The Problem posed in Section 2.1 requires us to prescribe ane i at inteb i for all vertices, i.e., target curvatures K
gle sums
rior verticesmostly 0 unless we want a cone singularityand
i for boundary vertices. For example, to
boundary curvatures
achieve a rectangular layout, which has obvious advantages for
e i = for all boundary vertices but
texture packing, one sets
e i = /2 (Figure 3, left; see also the Stamp
four, which receive
part in Figure 14).
4.2
b at some vertices
Alternatively, we may prescribe angle sums
and fix u at others. In this case the unique solution can be
found by considering the fixed ui as constants and minimizing
E(u) as a function of the remaining variables. Since the value
b at vertices with fixed u is clearly irrelevant for the
assigned to
minimization we may leave it undefined at these vertices. With
some ui fixed, the problem is not scale-invariant anymore and
the reduced Hessian becomes positive definite.
An extreme example is to fix u for all boundary vertices. This
gives complete control over the length distortion along the boundary while leaving the boundary curvature free.
Natural Boundary Conditions One choice is to set ui = 0 for
4.3
Topological Spheres
Meshes with sphere topology are sufficiently common that specific parameterization algorithms for them have been designed.
Figure 4: Sphere parameterizations of Dragon and Hygeia visualized as spheres with the original normals.
Cone Singularities
Texture Rectification
Figure 9: A layout for the Venus with four cones of each. These
map to corners of a parallelogram (left side). Using a checker board
texture leads to artifacts (inset). These can be removed by rectifying
the layout with an additional transformation (right side).
values may violate the Gauss-Bonnet relation, for example, the
sum of quantized cone values may be too large. In that case the
cone angle that was rounded up the most has its quantized value
decreased (unless it is /2 already). Repeating this procedure if
necessary leads to a set of quantized values which minimize the
maximum error between orginal and quantized values while satisfying the Gauss-Bonnet relation. (An analogous procedure can
be applied if the initial sum is too small.) Using these now fixed
cones the corresponding conformally equivalent metric is solved
for and laid out (Section 3.3). It defines the closed meta-polygon
P = {pi C | i = 0, . . . , m 1} given by the vertex positions
treated as complex numbersof m special vertices encountered
in a single, ordered traversal of the layout boundary. Special vertices are (1) cone vertices, (2) interior vertices with more than
two cut edges incident on them, and (3) boundary vertices with
a cut edge incident on them.
Figure 10 shows a meta-polygon for the Lion dataset. Five free
cones were placed with the simultaneous optimization procedure, quantized to the nearest multiple of /2 (3/2 for all five
cones in this case), and the corresponding conformally equivalent
metric laid out.
Figure 10: Meta-polygon for the Lion with quantized cones (solid
red) before rectification. After rectification, which forces all cone
points to integer locations, the texture tiles the model seamlessly
(inset shows 3/2 cone on chin underside).
Figure 11: The Gargoylea topological spherewith eight optimized and subsequently quantized cones. On the top right the
optimized cone domain and in the center the result of quantizing
the cones. Since the cones do not have integer locations the checkerboard texture has seams.
6.1
6.2
When deforming the meta-polygon, associated pairs of metaedges must maintain their alignment through Euclidean motions
with rotations being multiples of /2. This requirement is best
treated by considering the meta-edge differences pi+1 pi =: di
C (all index arithmetic is understood modulo m). Meta-edges
due to cutting come in pairs (i, j) and we have
di = R i j d j
Figure 12: Rectified domain for Max Planck dataset and visualization of corresponding iso-lines on the mesh.
We have run comparisons between breadth first and Laplace
layout, and found the results to be of comparable accuracy even
very degenerate in many places and the trust region solver took
very small steps. As a quantitative distortion measure we used
the area weighted mean of quasi-conformality [Kharevych et al.
2006] (denoted kqck1 in the table), i.e., the ratio of larger to
smaller singular value of the surface to parameter plane tangent
map. A value of one indicates no angle distortion at all. While it
is difficult to compare these numbers across papers ours appear
generally somewhat worse than what was reported by Ben-Chen
et al. and Kharevych et al., for example. Given that we effectively
enforce more structure than Ben-Chen and co-workers, and do
not allow any approximation of the original data as Kharevych
and co-workers do, this is perhaps not surprising. The last column
in the table indicates the boundary conditions (free, rectangle,
sphere) or the number of optimized (o) free cones, whether they
were subsequently quantized (q), and/or rectified (r).
Model
|V | Iter. kukmax
Face
1042 16 1.164
Stamp
4100
6 0.431
Stamp
4100
6 0.742
Lion
8356
6 1.089
Olivier
24795
6 1.188
Max Planck 25445 11 0.908
Gargoyle
24990 56 1.849
Hygeia
140654
5 0.650
Torso
142348 12 1.957
Torso
142348 21 1.091
Dragon
152803
5 3.759
Julius
209083
5 1.173
Numerical Results
kqck1
cones
1.0702 rectangle
1.0541
free
1.0625 rectangle
1.0282
5 (oqr)
1.0087
18 (o)
1.0097
4 (oqr)
1.0509
8 (oq)
1.0120
sphere
1.0048
4 (o)
1.0036 14 (oqr)
1.0277
sphere
1.0027
free
Conclusion
References
Balay, S., Buschelman, K., Eijkhout, V., Gropp, W. D., Kaushik,
D., Knepley, M. G., McInnes, L. C., Smith, B. F., and Zhang, H.
2007. PETSc Users Manual. Tech. Rep. ANL-95/11 (Revision
2.3.3), Argonne National Laboratory. http://www.mcs.anl.
gov/petsc/.
Ben-Chen, M., Gotsman, C., and Bunin, G. 2008. Conformal
Flattening by Curvature Prescription and Metric Scaling. Comp.
Graph. Forum 27, 2, 449458.
Benson, S., McInnes, L. C., Mor, J., Munson, T., and Sarich,
J. 2007. TAO User Manual. Tech. Rep. ANL/MCS-TM-242
(Revision 1.9), Argonne National Laboratory. http://www.mcs.
anl.gov/tao.
Figure 14: Further examples of a free boundary (Julius) and use of
boundary curvatures to enforce a rectangular domain (Stamp).
Desbrun, M., Meyer, M., and Alliez, P. 2002. Intrinsic Parameterizations of Surface Meshes. Comp. Graph. Forum 21, 3,
209218.
Duffin, R. J. 1956. Basic Properties of Discrete Analytic Functions. Duke Math. J. 23, 335363.
and
Gradient of E
(13)
l
l
j
ijk+ log 2 sin i jk
ijk+ log ki j jk
ki+ log 2 sini j k jk
kij .
= 21
jk
ki
2 sin
ij
j
(the cir ijk ) = l ki /(2 sin
ki ) = l i j /(2 sin
kij ) = R
Since l jk /(2 sin
ijk +
ki +
kij ) = 0 (the angle sum is ), Eq. (13) follows.
jk (
ij,
jk ,
ki ) = +
One finds also (in view of Eq. (3)) ui f (
ij
j
ij,
jk ,
ki ) = (
k +
)/2, and hence
f (
ki
ij
ki
ij,
jk ,
ki )
ui f (
(ui
2
ijk .
+ u j + uk ) = 12
Hessian of E
To show Eq. (10) for the Hessian of E, note that (Hess E u)i is
the derivative of the ith component of grad E in the direction u.
Since
P
bi 1
ijk ,
(grad E)i = 12
t i jk 3vi
2
b i is constant, we are done once we have shown that the
and
ijk is
first-order change in
ijk =
1
2
ki (uk ui ) + 12 cot
kij (u j ui ).
cot
(14)
i j
ki )/2 = cot
kij
kij cot
ki
ki .
(
i j = ui + u j and
ki = uk + ui , one obtains
Since
j
kij
kij cot
ki
ki ,
(u j uk )/2 = cot
and two other such equations through cyclic permutation of
j
kij ,
ijk ,
ki , are lini jk. These three linear equations for
early dependent (they sum to zero). Two of them together with
j
ijk +
ki +
kij = 0 form a linear system of equations which
determines :
!
j
i
ijk cot
ki
cot
0
jkj 1 u j ui
A ki = 2 ui uk ,
A = cot i
.
kij
0
cot
jk
0
kij
Using the addition theorems for sine and cosine one shows
det A = 1. Applying Cramers rule then yields
!
j
1
ijk = det
ki
(u j ui ) cot
1
(ui uk )
2
kij
cot
Minimal Distortion
ij
ki )/2 = log sin
kij log sin
ki
(
Proof. In homogeneous coordinates w = (x, y, z), the circumcircle of 4 is described by the equation q(w) = 0 where q(w) is the
quadratic form q(w) = x 2 + y 2 + 2c xz + 2d yz + ez 2 with c, d, e
be
determined by the conditions q(w i ) = 0. Similarly, let q(w)
The circumthe quadratic form describing the circumcircle of 4.
if q(w) is up to some
circle of 4 is mapped to the cirumcircle of 4
constant factor identical to q( f (w)), i.e. q(w) = q( f (w)),
or equivalently, if the corresponding symmetric bilinear forms
w
0 ) satisfy b(w, w 0 ) = b( f (w), f (w 0 )). This is
b(w, w 0 ), b(w,
i, aj w
j ), because the w i form a
the case iff b(w i , w j ) = b(ai w
basis for R3 . But since
) dA = M u g(grad u
, N ) ds,
D = M g(grad u, grad u
where N is the outward pointing unit normal vector field on
ranges over all smooth harmonic functions,
the boundary. As u
, N ) ranges over all smooth functions h : M R satisg(grad u
fying
R
h ds = 0.
(15)
M