Top Banner
Planar Curve Evolution Ron Kimmel www.cs.technion.ac.il/ ~ron Computer Science Department Technion-Israel Institute of Technolog Geometric Image Processing Lab
25

Planar Curve Evolution Ron Kimmel ron Computer Science Department Technion-Israel Institute of Technology Geometric Image Processing.

Dec 21, 2015

Download

Documents

Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: Planar Curve Evolution Ron Kimmel ron Computer Science Department Technion-Israel Institute of Technology Geometric Image Processing.

Planar Curve EvolutionRon Kimmel

www.cs.technion.ac.il/~ron

Computer Science Department Technion-Israel Institute of Technology

Geometric Image Processing Lab

Page 2: Planar Curve Evolution Ron Kimmel ron Computer Science Department Technion-Israel Institute of Technology Geometric Image Processing.

Planar Curves C(p)={x(p),y(p)}, p [0,1]

y

x

C(0)

C(0.1) C(0.2)

C(0.4)

C(0.7)

C(0.95)

C(0.9)

C(0.8)

pC =tangent

Page 3: Planar Curve Evolution Ron Kimmel ron Computer Science Department Technion-Israel Institute of Technology Geometric Image Processing.

Arc-length and Curvature

s(p)= | |dp 0

p

| | 1,sC pC | |p

sp

CC

C

ssC N

1

ssC N

C

Page 4: Planar Curve Evolution Ron Kimmel ron Computer Science Department Technion-Israel Institute of Technology Geometric Image Processing.

Invariant arclength should be

1. Re-parameterization invariant

2. Invariant under the group of transformations

drCCCFdpCCCFw rrrppp ,...,,,...,,

Geometric measure

Transform

Page 5: Planar Curve Evolution Ron Kimmel ron Computer Science Department Technion-Israel Institute of Technology Geometric Image Processing.

Euclidean arclength

Length is preserved, thus ,

dpCdpdydxdp

dpdydxds pdp

dydp

dx 222222

ds dy

dx

1sC

dpCs p

L

ppp dsdpCCdpCL0

1

0

1

0

21

,Length Total

Page 6: Planar Curve Evolution Ron Kimmel ron Computer Science Department Technion-Israel Institute of Technology Geometric Image Processing.

Curvature flow

Euclidean geometric heat equation nCCC sssst

where

nCt

flow

Euclideantransform

Page 7: Planar Curve Evolution Ron Kimmel ron Computer Science Department Technion-Israel Institute of Technology Geometric Image Processing.

Curvature flow

Takes any simple curve into a circular point

in finite time proportional to the area inside

the curve Embedding is preserved (embedded

curves keep their order along the

evolution).

nCt

Gage-Hamilton

Grayson

Given any simple planar curve

First becomes convex

Vanish at aCircular point

Page 8: Planar Curve Evolution Ron Kimmel ron Computer Science Department Technion-Israel Institute of Technology Geometric Image Processing.

Important property

Tangential components do not affect the geometry of an evolving curve

nnVCVC tt

,

V

nnV

,

Page 9: Planar Curve Evolution Ron Kimmel ron Computer Science Department Technion-Israel Institute of Technology Geometric Image Processing.

1Area

Reminder: Equi-affine arclength

Area is preserved, thus

vC

vvC

dpCCv ppp

31

,

1, vvv CC

dsdsCCv sss

31

31

,

dsdv 31

re-parameterizationinvariance

Page 10: Planar Curve Evolution Ron Kimmel ron Computer Science Department Technion-Israel Institute of Technology Geometric Image Processing.

Affine heat equation

Special (equi-)affine heat flow

31

, where, nCnnCC vvvvt

nCt

31

Sapiro

Given any simple planar curve

First becomes convex

Vanish at anelliptical

pointflow

Affinetransform

Page 11: Planar Curve Evolution Ron Kimmel ron Computer Science Department Technion-Israel Institute of Technology Geometric Image Processing.

Constant flow

Offset curves Level sets of distance map Equal-height contours of

the distance transform Envelope of all disks of equal radius centered along the curve (Huygens principle)

nCt

Page 12: Planar Curve Evolution Ron Kimmel ron Computer Science Department Technion-Israel Institute of Technology Geometric Image Processing.

Constant flow

Offset curves

nCt

Change in topology

Shock

Cusp

Page 13: Planar Curve Evolution Ron Kimmel ron Computer Science Department Technion-Israel Institute of Technology Geometric Image Processing.

Area inside C

dpCCA p,21 Area is defined via

C

pC

Page 14: Planar Curve Evolution Ron Kimmel ron Computer Science Department Technion-Israel Institute of Technology Geometric Image Processing.

So far we defined

Constant flow Curvature flow Equi-affine flow

We would like to explore evolution properties of measures like curvature, length, and area

nCt

nCtnCt

31

1

0

21

, dpCCL pp dpCCA p,21

Page 15: Planar Curve Evolution Ron Kimmel ron Computer Science Department Technion-Israel Institute of Technology Geometric Image Processing.

For

L

ppp VdsdpCt

CdpCCt

dpCCt

At 0

21 ...,,,

nVCt

VV

CC

CC

tt ss

pp

ppp 2...,

,2

3

12

0

, , ...L

p p p p pL C C dp C C C dp Vdstt t

0

0

2

L

t

L

t

t ss

L Vds

A Vds

V V

Length

Area

Curvature

Page 16: Planar Curve Evolution Ron Kimmel ron Computer Science Department Technion-Israel Institute of Technology Geometric Image Processing.

Constant flow ( )1V

22

00

00

2

VV

LdsVdsA

dsVdsL

sst

LL

t

LL

tLength

Area

Curvature

The curve vanishes at 2)0(Lt

)0,(1)0,(),( pt

ptp Riccati eq.

Singularity (`shock’) at

)0,( pt

Page 17: Planar Curve Evolution Ron Kimmel ron Computer Science Department Technion-Israel Institute of Technology Geometric Image Processing.

Curvature flow ( )V

32

00

0

2

0

2

sssst

LL

t

LL

t

VV

dsVdsA

dsVdsLLength

Area

Curvature

The curve vanishes at 2)0(At

Page 18: Planar Curve Evolution Ron Kimmel ron Computer Science Department Technion-Israel Institute of Technology Geometric Image Processing.

Equi-Affine flow ( )31V

37

35

32

31

34

292

312

00

00

ssssst

LL

t

LL

t

VV

dsVdsA

dsVdsLLength

Area

Curvature

Page 19: Planar Curve Evolution Ron Kimmel ron Computer Science Department Technion-Israel Institute of Technology Geometric Image Processing.

Geodesic active contours

nnyxgyxgCt

),,(),(

Goldenberg, Kimmel, Rivlin, Rudzsky,

IEEE T-IP 2001

Page 20: Planar Curve Evolution Ron Kimmel ron Computer Science Department Technion-Israel Institute of Technology Geometric Image Processing.

Tracking in color movies

Goldenberg, Kimmel, Rivlin, Rudzsky, IEEE T-IP 2001

nnyxgyxgCt

),,,(),,(

Page 21: Planar Curve Evolution Ron Kimmel ron Computer Science Department Technion-Israel Institute of Technology Geometric Image Processing.

From curve to surface evolution

It’s a bit more than invariant measures…

Page 22: Planar Curve Evolution Ron Kimmel ron Computer Science Department Technion-Israel Institute of Technology Geometric Image Processing.

Surface

A surface, For example, in 3D

Normal

Area element Total area

2 M: 2 nS nR

),(),,(),,(),( vuzvuyvuxvuS

vu

vu

SS

SSN

N

uS

vS

u vdA S S dudv

dudvSSA vu

Page 23: Planar Curve Evolution Ron Kimmel ron Computer Science Department Technion-Israel Institute of Technology Geometric Image Processing.

Surface evolution

Tangential velocity has no influence on the geometry

Mean curvature flow,area minimizing

NNVt

SV

t

S ,

NNV

, V

NHt

S

Page 24: Planar Curve Evolution Ron Kimmel ron Computer Science Department Technion-Israel Institute of Technology Geometric Image Processing.

Segmentation in 3D

Change in topology

Caselles,Kimmel, Sapiro, Sbert, IEEE T-PAMI 97

nnzyxgHzyxgSt

),,,(),,(

Page 25: Planar Curve Evolution Ron Kimmel ron Computer Science Department Technion-Israel Institute of Technology Geometric Image Processing.

Conclusions

Constant flow, geometric heat equations Euclidean Equi-affine Other data dependent flows

Surface evolution

www.cs.technion.ac.il/~ron