(Ebook) Metrics of curves in shape optimization and

analysis by Mennucci A. ISBN 9780769523729, 0769523722

Pdf Download

https://ebooknice.com/product/metrics-of-curves-in-shape-
optimization-and-analysis-2045964

★★★★★
4.7 out of 5.0 (24 reviews )

DOWNLOAD PDF

ebooknice.com
 (Ebook) Metrics of curves in shape optimization and analysis
by Mennucci A. ISBN 9780769523729, 0769523722 Pdf Download

EBOOK

Available Formats

■ PDF eBook Study Guide Ebook

EXCLUSIVE 2025 EDUCATIONAL COLLECTION - LIMITED TIME

INSTANT DOWNLOAD VIEW LIBRARY

We have selected some products that you may be interested in
Click the link to download now or visit ebooknice.com
for more options!.

(Ebook) Shape optimization and spectral theory by Antoine Henrot

ISBN 9783110550856, 3110550857

https://ebooknice.com/product/shape-optimization-and-spectral-
theory-11305268

(Ebook) Invariant Distances and Metrics in Complex Analysis by

Marek Jarnicki; Peter Pflug ISBN 9783110253863

https://ebooknice.com/product/invariant-distances-and-metrics-in-complex-
analysis-49193690

(Ebook) Shape Optimization and Spectral Theory by Antoine Henrot

ISBN 9783110550856, 9783110550887, 3110550857, 3110550881

https://ebooknice.com/product/shape-optimization-and-spectral-
theory-47486412

(Ebook) EXERGY, ENERGY SYSTEM ANALYSIS AND OPTIMIZATION:

Thermoeconomic Analysis Modeling, Simulation and Optimization in
Energy Systems by Editor: Christos A. Frangopoulos ISBN
9781848261655, 9781848266155, 9781848261648, 9781848266148,
9781848261662, 1848261659, 1848266154, 1848261640, 1848266146
https://ebooknice.com/product/exergy-energy-system-analysis-and-
optimization-thermoeconomic-analysis-modeling-simulation-and-optimization-
in-energy-systems-42594788
(Ebook) Invariant distances and metrics in complex analysis by
Jarnicki M., Pflug P. ISBN 9783110250435, 3110250438

https://ebooknice.com/product/invariant-distances-and-metrics-in-complex-
analysis-4582184

(Ebook) New Trends in Shape Optimization by Aldo Pratelli,

Günter Leugering (eds.) ISBN 9783319175621, 9783319175638,
3319175629, 3319175637

https://ebooknice.com/product/new-trends-in-shape-optimization-6744894

(Ebook) Applied shape optimization for fluids by Bijan

Mohammadi, Olivier Pironneau ISBN 9780199546909, 0199546908

https://ebooknice.com/product/applied-shape-optimization-for-fluids-1437342

(Ebook) Biota Grow 2C gather 2C cook by Loucas, Jason; Viles,

James ISBN 9781459699816, 9781743365571, 9781925268492,
1459699815, 1743365578, 1925268497

https://ebooknice.com/product/biota-grow-2c-gather-2c-cook-6661374

(Ebook) Software Metrics: A Guide to Planning, Analysis, and

Application by C. Ravindranath Pandian ISBN 9780203496077,
9780849316616, 0849316618, 0203496078

https://ebooknice.com/product/software-metrics-a-guide-to-planning-
analysis-and-application-1397834
 Metrics of curves in shape optimization and
analysis
Andrea C. G. Mennucci

Level set and PDE based reconstruction methods:

applications to inverse problems and image processing
CIME, Cetraro, 2008

Key words: Shape space, shape optimization, shape analysis, computer vision,
Riemannian geometry, manifold of curves, hyperspace of compact sets, edge
detection, image segmentation, visual tracking, curve evolution, gradient flow,
active contour, sobolev active contour.

Introduction
In these lecture notes we will explore the mathematics of the space of immersed
curves, as is nowadays used in applications in computer vision. In this field,
the space of curves is employed as a “shape space”; for this reason, we will also
define and study the space of geometric curves, that are immersed curves up
to reparameterizations. To develop the usages of this space, we will consider
the space of curves as an infinite dimensional differentiable manifold; we will
then deploy an effective form of calculus and analysis, comprising tools such
as a Riemannian metric, so as to be able to perform standard operations such
as minimizing a goal functional by gradient descent, or computing the distance
between two curves. Along this path of mathematics, we will also present some
current literature results. (Another common and interesting example of “shape
spaces” is the space of all compact subsets of lRn — we will briefly discuss this
option as well, and relate it to the aforementioned theory).
These lecture notes aim to be as self-contained as possible, so as to be acces-
sible to young mathematicians and non-mathematicians as well. For this reason,
many examples are intermixed with the definitions and proofs; in presenting
advanced and complex mathematical ideas, the rigorous mathematical definitions
and proofs were sometimes sacrificed and replaced with an intuitive description.
These lecture notes are organized as follows. Section 1 introduces the def-
initions and some basilar concepts related to immersed and geometric curves.
Section 2 overviews the realm of applications for a shape space in computer
vision, that we divide in the fields of “shape optimization” and “shape analysis”;

1
and hilights features and problems of those theories as were studied up to a
few years ago, so as to identify the needs and obstacles to further developments.
Section 3 contains a summary of all mathematical concepts that are needed for
the rest of the notes. Section 4 coalesces all the above in more precise defini-
tions of spaces of curves to be used as “shape spaces”, and sets mathematical
requirements and goals for applications in computer vision. Section 5 indexes
examples of “shape spaces” from the current literature, inserting it in a common
paradigm of “representation of shape”; some of this literature is then elaborated
upon in the following sections 6,7,8,9, containing two examples of metrics of
compact subsets of lRn , two examples of Finsler metrics of curves, two examples
of Riemannian metrics of curves “up to pose”, and four examples of Riemannian
metrics of immersed curves. The last such example is the family of Sobolev-type
Riemannian metrics of immersed curves, whose properties are studied in Section
10, with applications and numerical examples. Section 11 presents advanced
mathematical topics regarding the Riemannian spaces of immersed and geometric
curves.
I gratefully acknowledge that a part of the theory and many numerical
experiments exposited were developed in joint work with Prof. Yezzi (GaTech)
and Prof. Sundaramoorthi (UCLA); other numerical experiments were by A.
Duci and myself. I also deeply thank the organizers for inviting me to Cetraro
to give the lectures that were the basis for this lecture notes.

1 Shapes & curves

What is this course about? In the first two sections we begin by summarizing in
a simpler form the definitions, reviewing the goals, and presenting some useful
mathematical tools.

1.1 Shapes
A wide interest for the study of shape spaces arose in recent years, in particular
inside the computer vision community. Some examples of shape spaces are as
follows.
• The family of all collections of k points in lRn .
00
11
111
000
00
11
000
111
• The family of all non empty compact subsets of lRn . 00
11
000
111
00
11
000
111
00
11
000
111

• The family of all closed curves in lRn .

There are two different (but interconnected) fields of applications for a good
shape space in computer vision:

shape optimization where we want to find the shape that best satisfies a
design goal; a topic of interest in engineering at large;

2
shape analysis where we study a family of shapes for purposes of statistics,
(automatic) cataloging, probabilistic modeling, among others, and possibly
create an a-priori model for a better shape optimization.

1.2 Curves
S 1 = {x ∈ lR2 | |x| = 1} is the circle in the plane. It is the template for all
possible closed curves. (Open curves will be called paths, to avoid confusion).

Definition 1.1 (Classes of curves) • A C 1 curve is a continuously dif-

n
ferentiable map c : S → lR such that the derivative c0 (θ) := ∂θ c(θ) exists
1

at all points θ ∈ S 1 .
• An immersed curve is a C 1 curve c such that c0 (θ) 6= 0 at all points
θ ∈ S1.
c : S1 → c(S 1 )

7→

Note that, in our terminology , the “curve” is the function c, and not just the
image c(S 1 ) inside lRn .
Most of the theory following will be developed for curves in lRn , when this
does not complicate the math. We will call planar curves those whose image
is in lR2 .
The class of immersed curves is a differentiable manifold. For the purposes
of this introduction, we present a simple, intuitive definition.
Definition 1.2 The manifold of (parametric) curves M is the set of all
closed immersed curves. Suppose that c ∈ M , c : S 1 → lRn is a closed immersed
curve.
• A deformation of c is a function h : S 1 → lRn .

• The set of all such h is the tangent space Tc M of M at c.

• An infinitesimal deformation of the curve c0 in “direction” h will yield (on
first order) the curve c0 (u) + εh(u).

• A homotopy C connecting c0 to c1 is a continuous function h

C : [0, 1] × S 1 → S 1 such that c0 (θ) = C(0, θ) and c1 (θ) = c1
M c0
C(1, θ).
By defining γ(t) = C(t, ·) we can associate C to a path γ :
[0, 1] → M in the space of curves M , connecting c0 to c1 .

3
1.3 Geometric curves and functionals
Shapes are usually considered to be geometric objects. Representing a curve
using c : S 1 → lRn forces a choice of parameterization, that is not really part
of the concept of “shape”. To get rid of this, we first summarily present what
reparameterizations are. (We will provide more detailed definitions and properties
in Section 3.8).

Definition 1.3 Let Diff(S 1 ) be the family of diffeomorphisms of S 1 : all the

maps φ : S 1 → S 1 that are C 1 and invertible, and the inverse φ−1 is C 1 .

Diff(S 1 ) enjoys some important mathematical properties:

• Diff(S 1 ) is a group, and its group operation is the composition φ, ψ 7→

φ ◦ ψ.
• Diff(S 1 ) acts on M , the action is the right function composition c, φ 7→ c◦φ.
This action is a reparameterization of c.

Definition 1.4 The quotient space

B = M/Diff(S 1 )

is the space of curves up to reparameterization, also called geometric curves

in the following. Two parametric curves c1 , c2 ∈ M such that c1 = c2 ◦ φ are the
same geometric curve inside B.

B is mathematically defined as the set B = {[c]} of all equivalence classes

[c] of curves that are equal but for reparameterization,

[c] := {c ◦ φ for φ ∈ Diff(S 1 )}.

Remark 1.5 We may also consider the family Diff+ (S 1 ) of diffeomorphisms

with derivative φ0 > 0. Its action does not change the orientation of a curve. We
may then consider the quotient w.r.t Diff+ (S 1 ). The quotient space M/Diff+ (S 1 )
is the space of geometric oriented curves.

We can now define the geometric functionals.

Definition 1.6 A functional F (c) defined on curves will be called geometric

if it is invariant w.r.to reparameterization of c, that is, if F (c1 ) = F (c2 ) when
c1 = c2 ◦ φ.

In this case, F may be “projected” to B = M/Diff(S 1 ), that is, it may be

considered as a function F : B → lR.
It is important to remark that “geometric” theories have often provided the
best results in computer vision.

4
1.4 Curve–related quantities
A good way to specify the design goal for shape optimization is to define an
objective function (a.k.a. energy) F : M → lR that is minimum in the curve
that is most fit for the task.
When designing our F , we will want it to be geometric; this is easily ac-
complished if we use geometric quantities to start from. We now list the most
important such quantities.
In the following, given v, w ∈ lRn we will write |v| for the standard Euclidean
norm, and hv, wi or (v · w) for the standard scalar product. We will again
write c0 (θ) := ∂θ c(θ).
Definition 1.7 (Derivations) If the curve c is immersed, we can define the
derivation with respect to the arc parameter
1
∂s = ∂θ .
|c0 |
(We will sometimes also write Ds instead of ∂s .)
Definition 1.8 (Tangent) At all points where c0 (θ) 6= 0, we define the tangent
vector
c0 (θ)
T (θ) = 0 = ∂s c(θ) .
|c (θ)|
(At the points where c0 = 0 we may define T = 0).
It is easy to prove (and quite natural for our geometric intuition) that T is a
geometric quantity: if c̃ = c ◦ φ and T̃ is its tangent, then T̃ = T ◦ φ.
Definition 1.9 (Length) The length of the curve c is
Z
len(c) := |c0 (θ)| dθ . (1.9.∗)
S1

We can define formally the arc parameter s by

ds := |c0 (θ)| dθ ;
we use it only in integration, as follows.
Definition 1.10 (Integration by arc parameter) We define the integra-
tion by arc parameter of a function g : S 1 → lRk along the curve c by
Z Z
g(s) ds := g(θ)|c0 (θ)| dθ .
c S1

and the average integral

Z Z
1
g(s) ds := g(s) ds
c len(c) c

and we will sometimes shorten this as avgc (g).

5
 κ>0

H N

N
H

H
N
κ<0

Figure 1: Example of a regular curve and its curvature.

1.4.1 Curvature
Suppose moreover that the curve is immersed and is C 2 regular (that means that
it is twice differentiable, and the second derivative is continuous); in this case
we may define curvatures, that indicate how much the curve bends. There are
two different definitions of curvature of an immersed curve: mean curvature H
and signed curvature κ, that is defined when c is planar. H and k are extrinsic
curvatures, they are properties of the embedding of c into lRn .

Definition 1.11 (H) If c is C 2 regular and immersed, we can define the mean
curvature H of c as
H = ∂s ∂s c = ∂s T

where ∂s = |c10 | ∂θ is the derivation with respect to the arc parameter. It enjoys
the following properties.

Properties 1.12 • It is easy to prove that H ⊥ T .

• H is a geometric quantity. If c̃ = c ◦ φ and H̃ is its curvature, then
H̃ = H ◦ φ.
• 1/|H| is the radius of a tangent circle that best approximates the curve to
second order.

Definition 1.13 (N) When the curve c is immersed and planar, we can define
a normal vector N to the curve, by requiring that |N | = 1, N ⊥ T and N is
rotated π/2 degree anticlockwise with respect to T .

Definition 1.14 (κ) If c is planar and C 2 regular, then we can define a signed
scalar curvature κ = hH, N i, so that

∂s T = κN = H and ∂s N = −κT .

See fig. 1.

6
2 Shapes in applications
A number of methods have been proposed in shape analysis to define distances
between shapes, averages of shapes and statistical models of shapes. At the
same time, there has been much previous work in shape optimization, for ex-
ample image segmentation via active contours, 3D stereo reconstruction via
deformable surfaces; in these later methods, many authors have defined energy
functionals F (c) on curves (or on surfaces), whose minima represent the desired
segmentation/reconstruction; and then utilized the calculus of variations to derive
curve evolutions to search minima of F (c), often referring to these evolutions
as gradient flows. The reference to these flows as gradient flows implies a cer-
tain Riemannian metric on the space of curves; but this fact has been largely
overlooked. We call this metric H 0 , and properly define it in eqn. (2.9).

2.1 Shape analysis

Many method and tools comprise the shape analysis. We may list
• distances between shapes,
• averages for shapes,
• principle component analysis for shapes and
• probabilistic models of shapes.
We will present a short overview of the above, in theory and in applications. We
begin by defining the distance function and signed distance function, two tools
that we will use often in this theory.
Definition 2.1 Let A, B ⊂ lRn be compact.
• uA (x) := inf y∈A |x − y| is the distance function,

uA

uB

x2

B
A
x1

• bA (x) := uA (x) − ulRn \A (x) is the signed distance function.

bA

bB

x2

B
A
x1

7
2.1.1 Shape distances
A variety of distances have been proposed for measuring the difference between
two given shapes. Two examples follows.
Definition 2.2 The Hausdorff distance
   
dH (A, B) := sup uB (x) ∨ sup uA (x) (2.2.∗)
x∈A x∈B

where A, B ⊂ lRn are compact, and uA (x) is the distance function.

The above can be used for curves; in this case a distance may be defined by
associating to the two curves c1 , c2 the Hausdorff distance of the image of the
curves
dH (c1 (S 1 ), c2 (S 1 )) .

If c1 , c2 are immersed and planar, we may otherwise use dH (c̊1 ,c̊2 ) where
c̊1 ,c̊2 denote the internal region enclosed by c1 , c2 .

Definition 2.3 Let A, B be two measurable sets, let

A∆B := (A \ B) ∪ (A \ B)

be the set symmetric difference; let |A| be the Lebesgue measure of A. We

define the set symmetric distance as

d(A, B) = A∆B .

In the case of planar curves c1 , c2 , we can apply to above idea to define

d(c1 , c2 ) = c̊1 ∆c̊2 .

2.1.2 Shape averages

Many definitions have also been proposed for the average c̄ of a finite set of
shapes c1 , . . . , cN . There are methods based on the arithmetic mean (that are
representation dependent).
• One such case is when the shapes are defined by a finite family of N
parameters; so we can define the parametric or landmark averaging
N
1 X
c̄(p) = cn (p)
N n=1

where p is a common parameter/landmark.

8
 • More in general, we can define the signed distance level set averaging
by using as a representative of a shape its signed distance function, and
computing the average shape by
N
 1 X
c̄ = x | b̄(x) = 0 , where b̄(x) = bc (p)
N n=1 n

where bcn is the signed distance function of cn ; or in case of planar curves,

of the part of plane enclosed by cn .
Then there are non parametric, representation independent, methods. The
(possibly) most famous one is the
Definition 2.4 The distance-based average.1 Given a distance dM between
shapes, an average shape c̄ may be found by minimizing the sum of its squared
distances.
N
X
c̄ = arg min dM (c, cn )2
c
n=1

It is interesting to note that in Euclidean spaces there is an unique minimum,

that coincides with the arithmetic mean; while in general there may be no
minimum, or more than one.

2.1.3 Principal component analysis (PCA)

Definition 2.5 Suppose that X is a random vector taking values in lRn ; let
X = E(X) be the mean of X. The principal component analysis is the
representation of X as
Xn
X=X+ Yi Si
i=1

where Si are constant vectors, and Yi are uncorrelated real random variables
with zero mean, and with decreasing variance. Si is known as the i-th mode of
principal variation.

The PCA is possible in general in any finite dimensional linear space equipped
with an inner product. In infinite dimensional spaces, or equivalently in case of
random processes, the PCA is obtained by the Karhunen-Loève theorem.
Given a finite number of samples, it is also possible to define an empirical
principal component analysis, by estimating the expectation and variance of the
data.
In many practical cases, the variances of Yi decrease quite fast: it is then
sensible to replace X by a simplified variable
k
X
X̃ := X + Yi Si
i=1
1 Due to Fréchet, 1948; but also attributed to Karcher.

9
with k < n.
So, if shapes are represented by some “linear” representation, the PCA is
a valuable tool to study the statistics, and it has then become popular for
imposing shape priors in various shape optimization problems. For more general
manifold representations of shapes, we may use the exponential map (defined in
3.30), replacing Sn by tangent vectors and following geodesics to perform the
summation.
We present here an example in applications.

Example 2.6 (PCA by signed distance representation) In these pictures

we see an example of synergy between analysis and optimization, from Tsai et al.
[60]. The figure 2 contains a row of pictures that represent the (empirical)
PCA of signed distance function of a family of plane shapes. The figure 3
contains a second row of images that represent: an image of a plane shape (a)
occluded (b) and noise degraded (c), an initialization for the shape optimizer (d)
and the final optimal segmentation (e).

(a) (b) (c) (d) (e) (f) (g)

| {z } | {z } | {z } | {z }
mean first mode second mode third mode

Figure 2: PCA of plane shapes. (From Tsai et al. [60] c 2001 IEEE. Reproduced with
permission).

(a) (b) (c) (d) (e)

Figure 3: Segmentation of occluded plane shape. (From Tsai et al. [60] c 2001 IEEE.
Reproduced with permission).

10
2.2 Shape optimization & active contours
2.2.1 A short history of active contours
In the late 20th century, the most common approach to image segmentation
was a combination of edge detection and contour continuation. With edge
detection [5], small edge elements would be identified by a local analysis of
the image; then a continuation method would be employed to reconstruct
full contours. The methods were suffering from two drawbacks, edge detection
being too sensitive to noise and to photometric features (such as, sharp shadows,
reflections) that were not related to the physical structure of the image; and
continuation was in essence a NP-complete algorithm.
Active contours, introduced by Kass et al. [26], have been widely used for
the segmentation problem. The idea is to minimize an energy F (c) (where the
variable c is a contour i.e. a curve), that contains an edge-based attraction
term and a smoothness term, which becomes large when the curve is irregular.
An evolution is derived to minimize the energy based on principles from the
calculus of variations.
An unjustified feature of the model of [26] is that the evolution is dependent
on the way the contour is parameterized. Hence there have been geometric
evolutions similar to the idea of [26] in Caselles et al. [6], Malladi et al. [33],
which can be implemented by the level set method by Osher and Sethian [44].
Thereafter, Kichenassamy et al. [27] Caselles et al. [7] considered minimizing
a geometric energy, which is a generalization of Euclidean arc length, defined
on curves for the edge detection problem. The authors derived the gradient
descent flow in order to minimize the geometric energy.
In contrast to the edge-based approaches for active contours (mentioned
above), region-based energies for active contours have been proposed in Ronfard
[46] Zhu et al. [69] Yezzi et al. [64] Chan and Vese [8]. In these approaches, an
energy is designed to be minimized when the curve partitions the image into
statistically distinct regions. This kind of energy has provided many desirable
features; for example, it provides less sensitivity to noise, better ability to capture
concavities of objects, more dependence on global features of the image, and less
sensitivity to the initial contour placement.
In Mumford and Shah [42, 43], the authors introduce and rigorously study a
region-based energy that is designed to both extract the boundary of distinct
regions while also smoothing the image within these regions. Subsequently,
Tsai et al. [61] Vese and Chan [63] gave a curve evolution implementation of
minimizing the energy functional considered by Mumford&Shah in a level set
method; the gradient descent flows are calculated to minimize these energies.

2.2.2 Energies in computer vision

A variational approach to solving shape optimization problems is to define and
consequently minimize geometric energy functionals F (c) where c is a planar
curve. We may identify two main families of energies,

11
 • the region based energies:
Z Z
R2
F (c) = fin (x) dx + fout (x) dx
R Rc

where · · · dx is area element integral, and R and Rc

R
c
are the interior and exterior areas outlined by c; and
Rc R
• the boundary based energies:
Z
F (c) = φ(c(s)) ds
c R
where φ is designed to be small on the salient features of the image, and C · · · ds
is the integration by arc parameter defined in 1.10. Note that φ(x)ds may be
interpreted as a conformal metric on lR2 , so minima curves are geodesics w.r.to
the conformal metric.

2.2.3 Examples of geometric energy functionals for segmentation

Suppose I : Ω → [0, 1] is the image (in black & white). We propose two examples
of energies taken from the above families.
• The Chan-Vese segmentation energy [63]

Z 2
Z 2
F (c) = I(x) − avgR (I) dx + I(x) − avgRc (I) dx
R Rc
Z
1
where avgR I = I(x)dx is the average of I on the region R.
|R| R

• The geodesic active contour, (Caselles et al. [7],Kichenassamy et al. [27])

Z
F (c) = φ(c(s)) ds (2.7)
c
where φ may be chosen to be
1
φ(x) =
1 + |∇I(x)|2

that is small on sharp discontinuities of I (∇I is the gradient of I w.r.to

x). Since real images are noisy, the function φ in practice would be more
influenced by the noise than by the image features; for this reason, usually
the function φ is actually defined by
1
φ(x) =
1 + |∇G ? I(x)|2

where G is a smoothing kernel, such as the Gaussian.

12
2.3 Geodesic active contour method
2.3.1 The “geodesic active contour” paradigm
The general procedure for geodesic active contours goes as follows:
1. Choose an appropriate geometric energy functional, E.
2. Compute the directional derivative (a.k.a Gâteaux differential)2
d
DE(c)(h) = E(c + th)|t=0
dt
where c is a curve and h is an arbitrary perturbation of c.
3. Manipulate DE(c)(h) into the form
Z
h(s) · v(s) ds .
c

4. Consider v to be the “gradient”, the direction which increases E fastest.

5. Evolve c = c(t, θ) by the differential equation ∂t c = −v; this is called the
gradient descent flow.

2.3.2 Example: geodesic active contour edge model

We propose an explicit computation starting from the classical active contour
model.
1. Start from an energy that is minimal along image contours:
Z
E(c) = φ(c(s)) ds
c
where φ is defined to extract relevant features; for examples as discussed
after eqn. (2.7).
2. Calculate the directional derivative:
Z
DE(c)(h) = ∇φ(c) · h + φ(c)(Ds h · Ds c) ds (2.8)
Zc


= h · − φκN + (∇φ · N )N ds .
c

3. Deduce the “gradient”:

∇E = −φκN + (∇φ · N )N .

4. Write the gradient flow

∂c
= φκN − (∇φ · N )N .
∂t
(Note that the flow of geometric energies moves only in orthogonal direction w.r.t
the curve — this will be properly explained in in Section 11.10.)
2A discussion of all this will be in Section 3.7

13
2.3.3 Implicit assumption of H 0 inner product
We have made a critical assumption in going from the directional derivative
Z
DE(c)(h) = h(s) · v(s) ds
c

to deducing that the gradient of E is ∇E(c) = v. Namely, the definition of the

gradient 3 is based on the following equality
Z
hh(s), ∇Ei = h(s) · v(s) ds,
c |{z} |{z}
h1 =h h2 =∇E

that needs an inner-product structure.

This implies that we have been presuming all along that curves are equipped
with a H 0 -type inner-product defined as follows
Z
h1 , h2 H 0 := h1 (s) · h2 (s) ds . (2.9)
c

2.4 Problems & goals

We will now briefly list some examples that show some limits of the usual active
contour method.

2.4.1 Example: geometric heat flow

We first review one of the most mathematically studied gradient descent flows of
curves. By direct computation, the Gâteaux differential of the length of a closed
curve is Z Z
∂ len(c)
= h∂s h · T i ds = − hh · Hi ds (2.10)
∂h S1 S1

Let C = C(t, θ) be an evolving family of curves. The geometric heat flow

(also known as motion by mean curvature) is

∂C
= ∂s ∂s C
∂t
In the H 0 inner-product, this is the gradient descent for curve length.4
This flow has been studied deeply by the mathematical community, and is
known to enjoy the following properties.
Properties 2.11 • Embedded planar curves remain embedded.
• Embedded planar curves converge to a circular point.

3 The precise definition of what the gradient is is in section 3.7.

4A different gradient descent flow for curve length will be discussed in 10.45.

14
 • Comparison principle: if two embedded curves are used as initialization,
and the first contains the second, then it will contain the second for all
time of evolution. This is important for level set methods.
• The flow is well posed only for positive times, that is, for increasing t
(similarly to the usual heat equation).

For the proofs, see Gage and Hamilton [21], Grayson [23].

2.4.2 Short length bias

The usual edge-based active contour energy is of the form
Z
E(c) = g(c(s)) ds
c

supposing that g is reasonably constant in the region where c is running, then

E(c) ' g len(c), due to the integral being performed w.r.to the arc parameter ds.
So one way to reduce E is to shrink the curve. Consequently, when the image is
smooth and featureless (as in medical imaging), the usual edge based energies
would often drive the curve to a point.
To avoid it, an inflationary term νN (with ν > 0) was usually added to
the curve evolution, to obtain
∂c
= φκN − ∇φ + νN ,
∂t
see [7, 27]. Unfortunately, this adds a parameter that has to be tuned to match
the characteristics of each specific image.

2.4.3 Average weighted length

As an alternative approach we may normalize the energy to obtain a more
length-independent energy, the average weighted length
Z
1
avgc (g(c)) := g(c(s)) ds (2.12)
len(c) c

but this generates an ill-posed H 0 -flow, as we will now show.

2.4.4 Flow computations

Here we write L := len(c). Let g : lRn → lRk ; let
Z Z
1 1
avgc (g(c)) := g(c(s)) ds = g(c(θ))|c0 (θ)| dθ ;
L c L S1

15
then the Gâteaux differential of avgc (g(c)) is
Z
∂avgc (g(c)) 1
= ∇g(c) · h + g(c)h∂s h · T i ds −
∂h L c
Z Z
1
− 2 g(c) ds h∂s h · T i ds =
L c c
Z


= ∇g(c) · h + g(c) − avgc (g(c)) h∂s h · T i ds . (2.13)
c

In the above, we omit the argument (s) for brevity; ∇g is the gradient of g w.r.to
x.
If the curve is planar, we define the N and T as by 1.13 and 1.14; then,
integrating by parts, the above becomes
Z
∂avgc (g(c)) 1
= ∇g(c) · h − h∇g(c) · T ihh · T i −
∂h L S1
− g(c) − avgc (g(c)) hh · Ds2 ci ds =


Z 


= ∇g(c) · N − κ g(c) − avgc (g(c)) hh · N i ds (.2.14)
S1

Suppose now that we have a shape optimization functional E including a term

of the form avgc (g(c)); let C = C(t, θ) be an evolving family of curves trying to
minimize E; this flow would contain a term of the form
∂C

= . . . g(c(s)) − avgc (g) κN . . .
∂t
unfortunately the above flow is ill defined: it is a backward-running geometric
heat flow on roughly half of the curve. We will come back to this example in
Section 10.8.1.

2.4.5 Centroid energy

We will now propose another simple example where the above phenomenon is
again evident.

Example 2.15 (Centroid-based energy) Let us fix a target point v ∈ lRn .

We recall that avgc (c) is the center of mass of the curve. The energy
1
E(c) := |avgc (c) − v|2 (2.15.∗)
2
penalizes the distance from the center of mass to v. Let in the following c =
avgc (c) for simplicity. The directional derivative of E in direction h is
D E
DE(c)(h) = c − v, D(c)(h)

16
where in turn (by eqn. (2.13) and (2.14))
Z
D(c)(h) = h + (c − c)(Ds h · Ds c) ds = (2.15.†)
Z
= h − Ds c (h · Ds c) − (c − c)(h · Ds2 c) ds

supposing that the curve is planar, then

h − Ds c (h · Ds c) = N (h · N )

so Z
DE(c)(h) = hc − v, N i(h · N ) − hc − v, c − ciκ(h · N ) ds .

The H 0 gradient descent flow is then

∂c
= −∇H 0 E(c) = h(v − c), N iN − κN (c − c), (v − c) .
∂t
P
The first term h(v − c) · N iN in this gradient descent v
flow moves the whole curve towards v.
Let P := {w : h(w − c) · (v − c) ≥ 0} be the half plane
“on the v side” . The second term −κN (c − c) · (v − c) c
in this gradient descent flow tries to decrease the curve c
length out of P and increase the curve length in P , and
this is ill posed.

We will come back to this example in Proposition 10.30.

2.4.6 Conclusions
More recent works that use active contours for segmentation are not only based
on information from the image to be segmented (edge-based or region-based),
but also prior information, that is information known about the shape of the
desired object to be segmented. The work of Leventon et al. [31] showed how to
incorporate prior information into the active contour paradigm. Subsequently,
there have been a number of works, for example Tsai et al. [62], Rousson and
Paragios [47], Chen et al. [12], Cremers and Soatto [13], Raviv et al. [45], which
design energy functionals that incorporate prior shape information of the desired
object. In these works, the main idea is designing a novel term of the energy
that is small when the curve is close, in some sense, to a pre-specified shape.
The need for prior information terms arose from several factors such as
• the fact that some images contain limited information,
• the energies functions considered previously could not incorporate complex
information,

17
 • the energies had too many extraneous local minima, and the gradient flows
to minimize these energies allowed for arbitrary deformations that gave
rise to unlikely shapes; as in the example in Fig. 4 of segmentation using
the Chan-Vese energy, where the flowing contour gets trapped into noise.

Figure 4: H 0 gradient descent flow of Chan–Vese energy, to segment a noisy

square

Works on incorporating prior shape knowledge into active contours have

led to a fundamental question on how to define distance between two curves
or shapes. Many works, for example Younes [68], Soatto and Yezzi [51], Mio
and Srivastava [40], Charpiat et al. [9], Michor and Mumford [37], Yezzi and
Mennucci [67, 65], in the shape analysis literature have proposed different ways
of defining this distance.
However, [37, 65] observed that all previous works on geometric active
contours that derive gradient flows to minimize energies, which were described
earlier, imply a natural notion of Riemannian metric, given by the geometric
inner product H 0 (that we just “discovered” in eqn. (2.9)). Subsequently, [37, 67]
have shown a surprising property: the H 0 Riemannian metric on the space of
curves is pathological, since the “distance” between any two curves is zero.
In addition to the pathologies of the Riemannian structure induced by H 0 ,
there are also undesirable features of H 0 gradient flows. Some of these features
are listed below.

1. There are no regularity terms in the definition of the H 0 inner product.

That is, there is nothing in the definition of H 0 that discourages flows that
are not smooth in the space of curves. Thus, when energies are designed
to depend on the image that is to be segmented, the H 0 gradient is very
sensitive to noise in the image.
Therefore, in geometric active contour models, a penalty on the curve’s
length is added to keep the curve smooth. However, this changes the
energy that is being optimized and ends up solving a different problem.
2. H 0 gradients, evaluated at a particular point on the curve, depend locally
on derivatives of the curve. Therefore, as the curve becomes non-smooth,
as mentioned above, the derivative estimates become inaccurate, and thus,
the curve evolution becomes inaccurate. Moreover, for region-based and
edge-based active contours, the H 0 gradient at a particular point on the
curve depends locally on image data at the particular point. Although
region-based energies may depend on global statistics, such as mean values

18
 of regions, the H 0 gradient still depends on local data. These facts imply
that the H 0 gradient descent flow is sensitive to noise and local features.
3. The H 0 norm gives non-preferential treatment to arbitrary deformations
regardless of whether the deformations are global motions (not changing
the shape of the curve) such as translations, rotations and scales or whether
they are more local deformations.
4. Many geometric active contours (such as edge and region-based active
contours) require that the unit normal to the evolving curve be defined. As
such, the evolution does not make sense for polygons. Moreover, since in
general, an H 0 active contour does not remain smooth, one needs special
numerical schemes based on viscosity theory in a level set framework to
define the flow.
5. Many simple and apparently sound energies cannot be implemented for
shape optimization tasks;
• some energies generate ill-posed H 0 flows;
• if an energy integrand uses derivatives of the curve of degree up to d,
then the PDE driving the flow has degree 2d; but derivatives of the
curves of high degree are noise sensitive and are difficult to implement
numerically, and
• the active contours method works best when implemented using the
level set method, but this is difficult for flows PDEs of degree higher
than 2.
In conclusion, if one wishes to have a consistent view of the geometry of the
space of curves in both shape optimization and shape analysis, then one should
use the H 0 metric when computing distances, averages and morphs between
shapes. Unfortunately, H 0 does not yield a well define metric structure, since
the associated distance is identically zero. So to achieve our goal, we will need
to devise new metrics.

3 Basic mathematical notions

In this section we provide the mathematical theory that will be needed in the
rest of the course. (Some of the definitions are usually known to mathematics’
students; we will present them nonetheless as a chance to remark less known
facts.)We will though avoid technicalities in the definitions, and for the most part
just provide a base intuition of the concepts. The interested reader may obtain
more details from a books in analysis, such as [2], in functional analysis, such
as [48], and in differential and Riemannian geometry, such as [14], [29] or [30].
We start with a basic notion, in a quite simplified form.
Definition 3.1 (Topological spaces) A topological space is a set M with
associated a topology τ of subsets, that are the open sets in M .

19
The topology is the simplest and most general way to define what are the
“convergent sequences of points” and the “continuous functions”. We will not
provide details regarding topological spaces, since in the following we will mostly
deal with normed spaces and metric spaces, where the topology is induced by a
norm or a metric. We just recall this definition.

Definition 3.2 A homeomorphism is an invertible continuous function φ :

M → N between two topological spaces M, N , whose inverse φ−1 is again
continuous.

3.1 Distance and metric space

Definition 3.3 Given a set M , a distance d = dM is a function

d : M × M → [0, ∞]

such that
1. d(x, x) = 0,

2. if d(x, y) = 0 then x = y,
3. d(x, y) = d(y, x) (d is symmetric)
4. d(x, z) ≤ d(x, y) + d(y, z) (the triangular inequality).

There are some possible generalizations.

• The second request may be waived, in this case d is a semidistance.

• The third request may be waived: then d would be an asymmetric

distance. Most theorems we will see can be generalized to asymmetric
distances; see [34].

3.1.1 Metric space

The pair (M, d) is a metric space. A metric space has a distinguished topology,
generated by balls of the form B(x, r) := {y | d(x, y) < r}; according to this
topology, xn →n x iff d(xn , x) →n 0; and functions are continuous if they map
convergent sequences to convergent sequences. We will assume in the following
that the reader is acquainted with the concepts of “open, closed, compact sets”
and “continuous functions” in metric spaces. We recall though the definition of
“completeness”.

Definition 3.4 A metric space (M, d) is complete iff, for any given sequence
(cn ) ⊂ M , the fact that
lim d(cm , cn ) = 0
m,n→∞

implies that cn converges.

20
Example 3.5 • lRn is usually equipped with the Euclidean distance
v
u n
uX
d(x, y) = |x − y| = t (xi − yi )2 ;
i=1

and (lRn , d) is a complete metric space.

• Any closed subset of a complete space is complete.
• If we cut away an accumulation point out of a space, the resulting space is
not complete.

A complete metric space is a space without “missing points”. This is important

in optimization: if a space is not complete, any optimization method that moves
the variable and searches for the optimal solution may fail since the solution
may, in a sense, be “missing” from the space.

3.2 Banach, Hilbert and Fréchet spaces

Definition 3.6 Given a vector space E, a norm k · k is a function

k · k : E → [0, ∞]

such that
1. k · k is convex

2. if kxk = 0 then x = 0.
3. kλxk = |λ| kxk for λ ∈ lR

Again, there are some possible generalizations.

• If the second request is waived, then k · k is a seminorm.
• If the third request holds only for λ ≥ 0, then the norm is asymmetric;
in this case, it may happen that kxk =6 k − xk.
Each (semi/asymmetric)norm k · k defines a (semi/asymmetric)distance

d(x, y) := kx − yk.

So a norm induces a topology.

21
3.2.1 Examples of spaces of functions
We present some examples and definitions.
Definition 3.7 A locally-convex topological vector space E (shortened
as l.c.t.v.s. in the following) is a vector space equipped with a collection of
seminorms k · kk (with k ∈ K an index set); the seminorms induce a topology,
such that cn → c iff kcn − ckk →n 0 for all k; and all vector space operations
are continuous w.r.to this topology.
The simplest example of l.c.t.v.s. is obtained when there is only one norm; this
gives raise to two renowned examples of spaces.
Definition 3.8 (Banach and Hilbert spaces) • A Banach space is a
vector space E with a norm k · k defining a distance d(x, y) := kx − yk such
that E is metrically complete.
• A Hilbert space
p is a space with an inner product hf, gi, that defines a
norm kf k := hf, gi such that E is metrically complete.
(Note that a Hilbert space is also a Banach space).

Example 3.9 Let I ⊂ lRk be open; let p ∈ [1, ∞]. A standard example of
Banach space is the Lp space of functions f : I → lRn with norm
sZ
kf kLp := p
|f (x)|p dx for p ∈ [1, ∞) , kf kL∞ := sup |f (x)| ;
I I

those spaces contain all functions such that |f |p is Lebesgue integrable (resp.
f ∈ L∞ when |f | is bounded, on I \ N where N is a set of measure zero). If
p = 2, L2 is a Hilbert space by inner product
Z
hf, gi := f (x) · g(x) dx .
I

Note that, in these spaces, by definition, f = g iff the set {f 6= g} has zero
Lebesgue measure.

3.2.2 Fréchet space

The following citations [24] are referred to the first part of Hamilton’s 1982
survey on the Nash&Moser theorem.

Definition 3.10 A Fréchet space E is a complete Hausdorff metrizable l.c.t.v.s.;

where we define that the l.c.t.v.s. E is
complete when, for any sequence (cn ), the fact that

lim kcm − cn kk = 0
m,n→∞

for all k implies that cn converges;

22
Hausdorff when, for any given c, if kckk = 0 for all k then c = 0;
metrizable when there are countably many seminorms associated to E.

The reason for the last definition is that, if E is metrizable, then we can define a
distance
∞
X 2−k kx − ykk
d(x, y) :=
1 + kx − ykk
k=0

that generates the same topology as the family of seminorms k · kk ; and the vice
versa is true as well, see [48].

3.2.3 More examples of spaces of functions

Example 3.11 Let I ⊂ lRm be open and non empty.
• The Banach space C j (I → lRn ), with associated norm

kf k := sup sup |f (i) (t)| ;

i≤j t

(i)
where f (i) is the j-th derivative. In this space fn → f iff fn → f (i)
uniformly for all i ≤ j.

• The Sobolev space H j (I → lRn ), with scalar product

Z
hf, giH n := f (t) · g(t) + · · · + f (j) · g (j) dt
c

where f (j) is the j-th derivative5 .

• The Fréchet space of smooth functions C ∞ (I → lRn ).

The seminorms are
kf kk = sup |f (k) (x)|
x∈I

where f (k) is the k-th derivative. In this space, fn → f iff all the derivatives
(k)
fn converge uniformly to derivatives f (k) .
This last is one of the strongest topology between the topologies usually associated
to spaces of functions.

5 The derivatives are computed in distributional sense, and must exists as Lebesgue

integrable functions

23
3.2.4 Dual spaces.
Definition 3.12 Given a l.c.t.v.s. E, the dual space E ∗ is the space of all
linear functions L : E → lR.

If E is a Banach space, it is easy to see that E ∗ is again a Banach space, with

norm
kLkE ∗ := sup |Lx| .
kxkE ≤1

The biggest problem when dealing with Fréchet spaces, is that the dual of
a Fréchet space is not in general a Fréchet space, since it often fails to be
metrizable. (In most cases, the duals of Fréchet spaces are “quite wide” spaces;
a classical example being the dual elements of smooth functions, that are the
distributions). So given F, G Fréchet spaces, we cannot easily work with “the
space L(F, G) of linear functions between F and G”.
As a workaround, given an auxiliary space H, we will consider “indexed
families of linear maps” L : F × H → G, where L(·, h) is linear, and L is jointly
continuous; but we will not consider L as a map

h 7→ (f 7→ L(f, h))
(3.13)
H → L(F, G)

3.2.5 Derivatives
An example is the Gâteaux differential.
Definition 3.14 We say that a continuous map P : U → G, where F, G are
Fréchet spaces and U ⊂ F is open, is Gâteaux differentiable if for any h ∈ F
the limit
P (f + th) − P (f )
DP (f )(h) := lim (3.14.∗)
t→0 t
exists. The map DP (f ) : F → G is the Gâteaux differential.

Definition 3.15 We say that P is C 1 if DP : U × F → G exists and is jointly

continuous.

(This is weaker than what is usually required in Banach spaces).

The basics of the usual calculus hold.

Theorem 3.16 ([24, Thm. 3.2.5]) DP (f )(h) is linear in h.

Theorem 3.17 (Chain rule [24, Thm. 3.3.4]) If P, Q are C 1 then P ◦ Q is

C 1 and

D(P ◦ Q)(f )(h) = DQ(P (f )) DP (f )(h) .

Also, the implicit function theorem holds, in the form due to Nash&Moser:
see again [24] for details.

24
3.2.6 Troubles in calculus
But some important parts of calculus are instead lost.

Example 3.18 Suppose P : U ⊂ F → F is smooth, and consider the O.D.E.

d
f = Pf
dt
to be solved for a solution f : (−ε, ε) → F .

• if F is a Banach space, then, given initial condition f (0) = x, the solution

will exist and be unique (for ε > 0 small enough);
• but if F is a Fréchet space, then f may fail to exist or be unique.

See [24, Sec. 5.6]

A consequence (that we will discuss more later on) is that the exponential
map in Riemannian manifolds may fail to be locally surjective. See [24, Sec.
5.5.2].

3.3 Manifold
To build a manifold of curves, we have ahead two main definitions of “manifold”
to choose from.

Definition 3.19 (Differentiable Manifold) An abstract differentiable

manifold is a topological space M associated to a model l.c.t.v.s. U . It is
equipped with an atlas of charts φk : Uk → Vk , where Uk ⊂ U are open sets,
and Vk ⊂ M are open sets that cover M . The maps φk are homeomorphisms.
The composition φ−11 ◦ φ2 restricted to V1 ∩ V2 is usually V2
c
required to be a smooth map. V1
M
The dimension dim(M ) of M is the dimension of U . φ1 φ2

When M is finite-dimensional, the model space is always

U = lRn . U U1 U2

Since φk are homeomorphisms, then the topology of M is “identical” to the

topology of U . The rôle of the charts is to define the differentiable structure of
M mimicking that of U . See for example the definition of directional derivative
in eqn. (3.37.∗).

3.3.1 Submanifold
Definition 3.20 Suppose A, B are open subsets of two linear spaces. A diffeo-
morphism is an invertible differentiable function φ : A → B, whose inverse
φ−1 is again differentiable.

25
 Tc M

c
V1
M
φ1

x
U1 U

Figure 5: Tangent space

Let U be a fixed closed linear subspace of a l.c.t.v.s. X.

Definition 3.21 A submanifold is a subset M

c
of X, such that, at any point c ∈ M we may find M
V1

a chart φk : Uk → Vk , with Vk , Uk ⊂ X open sets, φ1

c ∈ Vk . The maps φk are diffeomorphisms, and X
φk maps U ∩ Uk onto M ∩ Vk .
U1 U

Most often, M = {Φ(c) = 0} where Φ : X → Y ; so, to prove that M is a

submanifold, we will use the implicit function theorem.
Note that M is itself an abstract manifold, and the model space is U ; so
dim(M ) = dim(U ) ≤ dim(X). Vice versa any abstract manifold is a submanifold
of some large X (by well known embedding theorems).

3.4 Tangent space and tangent bundle

Let us fix a differentiable manifold M , and c ∈ M . We want to define the
tangent space Tc M of M at c.
Definition 3.22 • The tangent space Tc M is more easily described for
submanifolds; in this case, we choose a chart φk and a point x ∈ Uk s.t.
φk (x) = c; Tc M is the image of the linear space U under the derivative
Dx φk . Tc M is itself a linear subspace in X. In figure 5, we graphically
represent Tc M though as an affine subspace, by translating it to the point
c.
• The tangent bundle T M is the collection of all tangent spaces. If M is
a submanifold of the vector space X, then

T M := {(c, h) | c ∈ M, h ∈ Tc M } ⊂ X × X .

The tangent bundle T M is itself a differentiable manifold; its charts are of the
form (φk , Dφk ), where φk are the charts for M .

3.5 Fréchet Manifold

When studying the space of all curves, we will deal with Fréchet manifolds,
where the model space E will be a Fréchet space, and the composition of local

26
charts φ−1
1 ◦ φ2 is smooth.
Some objects that we may find useful in the following are Fréchet manifolds.

Example 3.23 (Examples of Fréchet manifolds) Given two finite-dimensional

manifolds S, R, with S compact and n-dimensional,
• [24, Example 4.1.3]. the space C ∞ (S; R) of smooth maps f : S → R is a
Fréchet manifold. It is modeled on U = C ∞ (lRn ; R).
• [24, Example 4.4.6]. The space of smooth diffeomorphisms Diff(S) of S
onto itself is a Fréchet manifold. The group operation φ, ψ → φ ◦ ψ is
smooth.
• But if we try to model Diff(S) on C k (S; S), then the group operation is
not even differentiable.

• [24, Example 4.6.6]. The quotient of the two above

C ∞ (S; R)/Diff(S)

is a Fréchet manifold. It contains “all smooth maps from S to R, up to a

diffeomorphism of S”.

So the theory of Fréchet space seems apt to define and operate on the manifold
of geometric curves.

3.6 Riemann & Finsler geometry

We first define Riemannian geometries, and then we generalize to Finsler ge-
ometries.

3.6.1 Riemann metric, length

Definition 3.24 • A Riemannian metric G on a differentiable manifold
M defines a scalar product hh1 , h2 iG|c on h1 , h2 ∈ Tc M , dependent on
the point c ∈ M . We assume that the scalar product varies smoothly w.r.to
c.
p
• The scalar product defines the norm |h|c = |h|G|c = hh, hiG|c .

Suppose γ : [0, 1] → M is a path connecting c0 to c1 .

Z 1
• The length is Len(γ) := |γ̇(v)|γ(v) dv
0 γ
where γ̇(v) := ∂v γ(v). c1
M c0
Z 1
• The energy (or action) is E(γ) := |γ̇(v)|2γ(v) dv
0

27
3.6.2 Finsler metric, length
Definition 3.25 We define a Finsler metric to be a function F : T M → lR+ ,
such that
• F is continuous and,

• for all c ∈ M , v 7→ F (c, v) is a norm on Tc M .

We will sometimes write |v|c := F (c, v).

(Sometimes F is called a “Minkowsky norm”).

As for the case of norms, a Finsler metric may be asymmetric; but, for sake
of simplicity, we will only consider the symmetric case.
Using the norm |v|c we can then again define the length of paths by
Z 1 Z 1
LenF (γ) = |γ̇(t)|γ(t) dt = F (γ(t), γ̇(t)) dt
0 0

and similarly the action.

3.6.3 Distance
Definition 3.26 The distance d(c0 , c1 ) is the infimum of Len(γ) between all
C 1 paths γ connecting c0 , c1 ∈ M .

Remark 3.27 In the following chapter, we will define some differentiable man-
ifolds M of curves, and add a Riemann (or Finsler) metric G on those; there
are two different choices for the model space,
• suppose we model the differentiable manifold M on a Hilbert space U , with
scalar product h, iU ; this implies that M has a topology τ associated to
it, and this topology, through the charts φ, is the same as that of U . Let
now G be a Riemannian metric; since the derivative of a chart Dφ(c)
maps U onto Tc M , one natural hypothesis will be to assume that h, iU and
h, iG,c be locally equivalent (uniformly w.r.to small movements of c); as
a consequence, the topology generated by the Riemannian distance d will
coincide with the original topology τ . A similar request will hold for the
case of a Finsler metric G, in this case U will be a Banach space with a
norm equivalent to that defined by G on Tc M .

• We will though find out that, for technical reasons, we will initially model
the spaces of curves on the Fréchet space C ∞ ; but in this case there cannot
be a norm on Tc M that generates the same original topology (for the proof,
see I.1.46 in [48]).

28
3.6.4 Minimal geodesics
Definition 3.28 If there is a path γ ∗ providing the minimum of Len(γ) between
all paths connecting c0 , c1 ∈ M , then γ ∗ is called a minimal geodesic.

The minimal geodesic is also the minimum of the action (up to reparameteri-
zation).
Proposition 3.29 Let ξ ∗ provide the minimum minγ E(γ) in the class of all
paths γ in M connecting x to y. Then ξ ∗ is a minimal geodesic and its speed
|ξ˙∗ | is constant.
Vice versa, let γ ∗ be a minimal geodesic, then there is a reparameterization
ξ = γ ∗ ◦ φ s.t. ξ ∗ provides the minimum minγ E(γ).
∗

A proof may be found in [34].

3.6.5 Exponential map

The action E is a smooth integral, quadratic in γ̇, and we can compute the
Euler-Lagrange equations; its minima are more regular, since they are guaranteed
to have constant speed; consequently, when trying to find geodesics, we will try
to minimize the action and not the length. This also related to the idea of the
exponential map.

Definition 3.30 Let γ̈ = Γ(γ̇, γ) be the Euler-Lagrange ODE of the action

R1
E(γ) = 0 |γ̇(v)|2 dv. Any solution of this ODE is a critical geodesic. Note
that any minimal geodesic is a critical geodesic.
Define the exponential map expc : Tc M → M as expc (η) = γ(1), where
γ is the solution of
n
γ̈(v) = Γ(γ̇(v), γ(v)), γ(0) = c, γ̇(0) = η (3.30.∗)

Solving the above ODE (3.30.∗) is informally known as shooting geodesics.

The exponential map is often used as a chart, since it is the least possibly
deforming map at the origin.
Theorem 3.31 Suppose that M is a smooth differentiable manifold modeled
on a Hilbert space with a smooth Riemannian metric. The derivative of the
exponential map expc : Tc M → M at the origin is an isometry, hence expc is a
local diffeomorphism between a neighborhood of 0 ∈ Tc M and a neighborhood of
c ∈ M.
(See [30], VIII §5). The exponential map can then “linearize” small portions of
M , and so it will enable us to use linear methods such as the principal component
analysis. Unfortunately, the above result does not hold if M is modeled on a
Fréchet space.

29
3.6.6 Hopf–Rinow theorem
Theorem 3.32 (Hopf–Rinow) Suppose M is a finite dimensional Rieman-
nian or Finsler manifold. The following are equivalent:
• (M, d) is metrically complete;

• the O.D.E. (3.30.∗) can be solved for all c ∈ M , η ∈ Tc M and v ∈ lR;

• the map expc is surjective;
and all those imply that, ∀x, y ∈ M there exist a minimal geodesic connecting x
to y.

3.6.7 Drawbacks in infinite dimensions

In a certain sense, infinite dimensional manifolds are simpler than their corre-
sponding finite-dimensional counterparts: indeed, by Eells and Elworthy [18],
Theorem 3.33 (Eells–Elworthy) Any smooth differentiable manifold M mod-
eled on an infinite dimensional separable Hilbert space H may be embedded as
an open subset of that Hilbert space.
In other words, it is always possible to express M using one single chart. (But
note that this may not be the best way for computations/applications).
When M is infinite dimensional Riemannian manifold, though, only a small
part of the Hopf–Rinow theorem still holds.
Proposition 3.34 Suppose M is infinite dimensional, modeled on a Hilbert
space, and (M, d) is complete, then the O.D.E. (3.30.∗) of a critical geodesic
can be solved for all v ∈ lR.
But other implications fails.

Example 3.35 (Atkin [3]) There exists an infinite dimensional complete Hilbert
smooth manifold M and x, y ∈ M such that there is no critical geodesic connect-
ing x to y.

That is,
• (M, d) is complete 6⇒ expc is surjective,
• and (M, d) is complete 6⇒ minimal geodesics exist.

It is then, in general, quite difficult to prove that an infinite dimensional

manifold admits minimal geodesics (even when it is known to be metrically
complete). There are though some positive results, such as Ekeland [19] (that
we cannot properly discuss for lack of space); or the following.
Theorem 3.36 (Cartan–Hadamard) Suppose that M is connected, simply
connected and has seminegative curvature; then these are equivalent:

30
Other documents randomly have
different content
Welcome to our website – the ideal destination for book lovers and
knowledge seekers. With a mission to inspire endlessly, we offer a
vast collection of books, ranging from classic literary works to
specialized publications, self-development books, and children's
literature. Each book is a new journey of discovery, expanding
knowledge and enriching the soul of the reade

Our website is not just a platform for buying books, but a bridge
connecting readers to the timeless values of culture and wisdom. With
an elegant, user-friendly interface and an intelligent search system,
we are committed to providing a quick and convenient shopping
experience. Additionally, our special promotions and home delivery
services ensure that you save time and fully enjoy the joy of reading.

Let us accompany you on the journey of exploring knowledge and

personal growth!

ebooknice.com