Quasiconformal distortion of projective maps and discrete conformal - - PowerPoint PPT Presentation

quasiconformal distortion of projective maps and discrete
SMART_READER_LITE
LIVE PREVIEW

Quasiconformal distortion of projective maps and discrete conformal - - PowerPoint PPT Presentation

Quasiconformal distortion of projective maps and discrete conformal maps with Stefan Born and Ulrike B ucking arXiv:1505.01341 Bobenko, Pinkall, S Discrete conformal maps and ideal hyperbolic polyhedra Geom. Topol. 19-4 (2015), 2155-2215 S,


slide-1
SLIDE 1

Quasiconformal distortion of projective maps and discrete conformal maps

with Stefan Born and Ulrike B¨ ucking arXiv:1505.01341

Bobenko, Pinkall, S Discrete conformal maps and ideal hyperbolic polyhedra

  • Geom. Topol. 19-4 (2015), 2155-2215

S, Schr¨

  • der, Pinkall

Conformal equivalence of triangle meshes ACM Transactions on Graphics 27:3 (2008)

slide-2
SLIDE 2

1 Discrete conformal maps

slide-3
SLIDE 3

1 ✘✘✘✘✘

❳❳❳❳❳

Discrete conformal maps

slide-4
SLIDE 4

1 ✘✘✘✘✘

❳❳❳❳❳

Discrete conformal maps

◮ conformal means angle preserving ◮ lengths scaled by conformal factor

independent of direction dfp(v) = eu(p) v

◮ looks like a similarity

transformation when zooming in

slide-5
SLIDE 5

1 ✘✘✘✘✘

❳❳❳❳❳

Discrete conformal maps

◮ conformal means angle preserving ◮ lengths scaled by conformal factor

independent of direction dfp(v) = eu(p) v

◮ looks like a similarity

transformation when zooming in

slide-6
SLIDE 6

1 Discrete conformal maps: scale factors

Definition (Luo 2004)

Two triangulated surfaces are discretely conformally equivalent, if (i) triangulations are combinatorially equivalent (ii) edge lengths ℓij and ˜ ℓij related by ˜ ℓij = e

1 2 (ui+uj)ℓij

◮ Leads to rich theory with connections to hyperbolic geometry.

slide-7
SLIDE 7

1 Discrete conformal maps: length cross ratio

For interior edges ij define length cross ratio lcrij = ℓih ℓjk ℓhj ℓki

Theorem

ℓ, ˜ ℓ discretely conformally equivalent ⇐ ⇒ lcr = lcr

slide-8
SLIDE 8

1 Discrete conformal maps: Examples

slide-9
SLIDE 9

1 Discrete conformal maps: Examples

slide-10
SLIDE 10

1 Discrete conformal maps: Examples

slide-11
SLIDE 11

1 Discrete conformal maps: Examples

slide-12
SLIDE 12

1 Discrete conformal maps: Examples

slide-13
SLIDE 13

1 Discrete conformal maps: Examples

slide-14
SLIDE 14

1 Discrete conformal maps: Examples

slide-15
SLIDE 15

1 Discrete conformal maps: Examples

slide-16
SLIDE 16

1 Discrete conformal maps: Examples

slide-17
SLIDE 17

1 Discrete conformal maps: Interpolation

◮ How to interpolate over triangles? ◮ piecewise linear always works ◮ better: circumcircle preserving

piecewise projective (cpp) maps piecewise linear

slide-18
SLIDE 18

1 Discrete conformal maps: Interpolation

◮ How to interpolate over triangles? ◮ piecewise linear always works ◮ better: circumcircle preserving

piecewise projective (cpp) maps

slide-19
SLIDE 19

1 Discrete conformal maps: Interpolation

◮ How to interpolate over triangles? ◮ piecewise linear always works ◮ better: circumcircle preserving

piecewise projective (cpp) maps

slide-20
SLIDE 20

1 Discrete conformal maps: Interpolation

◮ How to interpolate over triangles? ◮ piecewise linear always works ◮ better: circumcircle preserving

piecewise projective (cpp) maps

A B C S A′ B′ C ′ S′

S, S′: symmedian (Lemoine, Grebe) points

slide-21
SLIDE 21

1 Discrete conformal maps: Interpolation

◮ How to interpolate over triangles? ◮ piecewise linear always works ◮ better: circumcircle preserving

piecewise projective (cpp) maps

Theorem

cpp maps fit together continuously across edges ⇐ ⇒ triangulations are discretely conformally equivalent

A B C S A′ B′ C ′ S′

S, S′: symmedian (Lemoine, Grebe) points

slide-22
SLIDE 22

1 Discrete conformal maps: Interpolation

◮ How to interpolate over triangles? ◮ piecewise linear always works ◮ better: circumcircle preserving

piecewise projective (cpp) maps

Definition

discrete conformal map: simplicial map, cpp on triangles

A B C S A′ B′ C ′ S′

S, S′: symmedian (Lemoine, Grebe) points

slide-23
SLIDE 23

1 Discrete conformal maps: Interpolation

◮ How to interpolate over triangles? ◮ piecewise linear always works ◮ better: circumcircle preserving

piecewise projective (cpp) maps

Definition

discrete conformal map: simplicial map, cpp on triangles

◮ cpp interpolation is “visibly smoother”

cpp

slide-24
SLIDE 24

1 Discrete conformal maps: Interpolation

◮ How to interpolate over triangles? ◮ piecewise linear always works ◮ better: circumcircle preserving

piecewise projective (cpp) maps

Definition

discrete conformal map: simplicial map, cpp on triangles

◮ cpp interpolation is “visibly smoother”

piecewise linear

slide-25
SLIDE 25

1 Discrete conformal maps: Interpolation

◮ How to interpolate over triangles? ◮ piecewise linear always works ◮ better: circumcircle preserving

piecewise projective (cpp) maps

Definition

discrete conformal map: simplicial map, cpp on triangles

◮ cpp interpolation is “visibly smoother”

linear

slide-26
SLIDE 26

1 Discrete conformal maps: Interpolation

◮ How to interpolate over triangles? ◮ piecewise linear always works ◮ better: circumcircle preserving

piecewise projective (cpp) maps

Definition

discrete conformal map: simplicial map, cpp on triangles

◮ cpp interpolation is “visibly smoother”

cpp

slide-27
SLIDE 27

1 Discrete conformal maps: Interpolation

◮ How to interpolate over triangles? ◮ piecewise linear always works ◮ better: circumcircle preserving

piecewise projective (cpp) maps

Definition

discrete conformal map: simplicial map, cpp on triangles

◮ cpp interpolation is “visibly smoother” ◮ Why? ◮ Lower quasiconformal distortion?

cpp

slide-28
SLIDE 28

2 Quasiconformal distortion

f z f (z) ε λ2ε λ1ε

◮ 0 ≤ λ2 ≤ λ1

singular values of dfz

◮ Df (z) = ±λ1

λ2 , sign depends on orientation

◮ |µf | = Df − 1

Df + 1 modulus of Beltrami differential

◮ |µf | =

     where df is conformal 1 where df is singular ∞ where df is anticonformal

slide-29
SLIDE 29

3 Distortion of a projective map

|µf | = 0 |µf | = ∞ |µf | = 1 on f −1(ℓ∞)

Theorem

(i) If projective map f : RP2 → RP2 is not affine, contourlines of |µf | form a hyperbolic pencil of circles. (ii) This hyperbolic pencil of circles is mapped to another hyperbolic pencil of circles.

slide-30
SLIDE 30

3 Distortion of a projective map

|µf | = 0 |µf | = ∞ |µf | = 1 on f −1(ℓ∞)

Corollary

If f is orientation preserving on triangle ABC, then max

z∈ABC |µf (z)| is attained at A,B, or C.

slide-31
SLIDE 31

3 Distortion of a projective map & circles mapped to circles

|µf | = 0 |µf | = ∞ |µf | = 1 on f −1(ℓ∞)

◮ Which circles are mapped to circles by a projective map f ?

slide-32
SLIDE 32

3 Distortion of a projective map & circles mapped to circles

|µf | = 0 |µf | = ∞ |µf | = 1 on f −1(ℓ∞)

◮ Which circles are mapped to circles by a projective map f ?

Theorem

◮ If f ∈ Sim: all circles ◮ If f ∈ Aff \ Sim: no circle ◮ If f ∈ Aff: exactly one hyperbolic pencil of circles

slide-33
SLIDE 33

4 Distortion of circumcircle preserving projective map

A B C S A′ B′ C ′ S′

Theorem

If f : ABC → A′B′C ′ is a cpp map, then |µf (A)| = |µf (B)| = |µf (C)| = |µh|, where h is the affine map ABC → A′B′C ′.

slide-34
SLIDE 34

4 Distortion of circumcircle preserving projective map

A B C S A′ B′ C ′ S′

Theorem

If f : ABC → A′B′C ′ is a cpp map, then |µf (A)| = |µf (B)| = |µf (C)| = |µh|, where h is the affine map ABC → A′B′C ′.

◮ cpp interpolation better than linear interpolation

(except at vertices)

slide-35
SLIDE 35

5 Angle bisector preserving projective map

A B C A′ B′ C ′

Theorem

Of all projective maps ABC → A′B′C ′, the angle bisector preserving projective map (app map) simultaneously minimizes |µf (A)|, |µf (B)|, |µf (C)|.

slide-36
SLIDE 36

5 Angle bisector preserving projective map

Theorem

Two triangulations are discretelely conformally equivalent ⇔ app maps are continuous across edges.

◮ follows from angle bisector theorem

a b p q

a b = p q

slide-37
SLIDE 37

5 Angle bisector preserving projective map

Which interpolation looks best? linear

slide-38
SLIDE 38

5 Angle bisector preserving projective map

Which interpolation looks best? app

slide-39
SLIDE 39

5 Angle bisector preserving projective map

Which interpolation looks best? cpp

slide-40
SLIDE 40

6 A 1-parameter familiy of projective interpolation schemes

◮ Barycenter has barycentric coordinates [1, 1, 1] ◮ Incircle center has barycentric coordinates [a, b, c] ◮ Symmedian point has barycentric coordinates [a2, b2, c2]

slide-41
SLIDE 41

6 A 1-parameter familiy of projective interpolation schemes

◮ Barycenter has barycentric coordinates [1, 1, 1] ◮ Incircle center has barycentric coordinates [a, b, c] ◮ Symmedian point has barycentric coordinates [a2, b2, c2] ◮ Exponent-t-center has barycentric coordinates [at, bt, ct]

slide-42
SLIDE 42

6 A 1-parameter familiy of projective interpolation schemes

◮ Barycenter has barycentric coordinates [1, 1, 1] ◮ Incircle center has barycentric coordinates [a, b, c] ◮ Symmedian point has barycentric coordinates [a2, b2, c2] ◮ Exponent-t-center has barycentric coordinates [at, bt, ct]

Theorem

The projective maps that map exponent-t-centers to exponent-t-centers fit together continuously across edges if, and for t = 0 only if, the triangulations are discretely conformally equivalent.

slide-43
SLIDE 43

6 A 1-parameter familiy of projective interpolation schemes

t = −1.0

slide-44
SLIDE 44

6 A 1-parameter familiy of projective interpolation schemes

t = −0.5

slide-45
SLIDE 45

6 A 1-parameter familiy of projective interpolation schemes

t = 0.0 (linear)

slide-46
SLIDE 46

6 A 1-parameter familiy of projective interpolation schemes

t = 0.5

slide-47
SLIDE 47

6 A 1-parameter familiy of projective interpolation schemes

t = 1.0 (app)

slide-48
SLIDE 48

6 A 1-parameter familiy of projective interpolation schemes

t = 1.5

slide-49
SLIDE 49

6 A 1-parameter familiy of projective interpolation schemes

t = 2.0 (cpp)

slide-50
SLIDE 50

6 A 1-parameter familiy of projective interpolation schemes

t = 2.5

slide-51
SLIDE 51

6 A 1-parameter familiy of projective interpolation schemes

t = 3.0