+ All Categories
Transcript
Page 1: Iterative Image Registration: Lucas & Kanade Revisited

Iterative Image Registration:

Lucas & Kanade Revisited

Kentaro Toyama

Vision Technology Group

Microsoft Research

Page 2: Iterative Image Registration: Lucas & Kanade Revisited

Every writer creates his own precursors. His work modifies our conception of the past, as it will modify the future.

Jorge Luis Borges

Page 3: Iterative Image Registration: Lucas & Kanade Revisited

History

• Lucas & Kanade (IUW 1981)

LK BAHH ST S BJ HB BL G SI CETSC

• Bergen, Anandan, Hanna, Hingorani (ECCV 1992)

• Shi & Tomasi (CVPR 1994)

• Szeliski & Coughlan (CVPR 1994)

• Szeliski (WACV 1994)

• Black & Jepson (ECCV 1996)

• Hager & Belhumeur (CVPR 1996)

• Bainbridge-Smith & Lane (IVC 1997)

• Gleicher (CVPR 1997)

• Sclaroff & Isidoro (ICCV 1998)

• Cootes, Edwards, & Taylor (ECCV 1998)

Page 4: Iterative Image Registration: Lucas & Kanade Revisited

Image Registration

Page 5: Iterative Image Registration: Lucas & Kanade Revisited

Applications

Page 6: Iterative Image Registration: Lucas & Kanade Revisited

Applications

• Stereo

LK BAHH ST S BJ HB BL G SI CETSC

Page 7: Iterative Image Registration: Lucas & Kanade Revisited

Applications

• Stereo

• Dense optic flow

LK BAHH ST S BJ HB BL G SI CETSC

Page 8: Iterative Image Registration: Lucas & Kanade Revisited

Applications

• Stereo

• Dense optic flow

• Image mosaics

LK BAHH ST S BJ HB BL G SI CETSC

Page 9: Iterative Image Registration: Lucas & Kanade Revisited

Applications

• Stereo

• Dense optic flow

• Image mosaics

• Tracking

LK BAHH ST S BJ HB BL G SI CETSC

Page 10: Iterative Image Registration: Lucas & Kanade Revisited

Applications

• Stereo

• Dense optic flow

• Image mosaics

• Tracking

• Recognition

LK BAHH ST S BJ HB BL G SI CETSC

?

Page 11: Iterative Image Registration: Lucas & Kanade Revisited

Lucas & Kanade

#1

Derivation

Page 12: Iterative Image Registration: Lucas & Kanade Revisited

L&K Derivation 1

I0(x)

)('0 xI

h

xIhxIh

)()(lim 00

0

)('0 xI

Page 13: Iterative Image Registration: Lucas & Kanade Revisited

L&K Derivation 1

)('0 xI

h

xIhxI )()( 00

h I0(x)

I0(x+h)

Page 14: Iterative Image Registration: Lucas & Kanade Revisited

L&K Derivation 1

h I0(x)

)('0 xI

h

xIxI )()( 0

I(x)

Page 15: Iterative Image Registration: Lucas & Kanade Revisited

L&K Derivation 1

h I0(x)

h)(

)()('0

0

xI

xIxI

I(x)

Page 16: Iterative Image Registration: Lucas & Kanade Revisited

L&K Derivation 1

I0(x)

h

Rx xI

xIxI

R )(

)()(

||

1'0

0

RI(x)

Page 17: Iterative Image Registration: Lucas & Kanade Revisited

L&K Derivation 1

I0(x)

h

RxxxI

xIxIxw

xw )(

)]()()[(

)(

1'0

0

I(x)

Page 18: Iterative Image Registration: Lucas & Kanade Revisited

L&K Derivation 1

h0 I0(x)

0h

I(x)

RxxxI

xIxIxw

xw )(

)]()()[(

)(

1'0

0

Page 19: Iterative Image Registration: Lucas & Kanade Revisited

L&K Derivation 1

1h

Rxx

hxI

hxIxIxw

xwh

)(

)]()()[(

)(

1

0'0

000

I0(x+h0)

I(x)

Page 20: Iterative Image Registration: Lucas & Kanade Revisited

L&K Derivation 1

2h

Rxx

hxI

hxIxIxw

xwh

)(

)]()()[(

)(

1

1'0

101

I0(x+h1)

I(x)

Page 21: Iterative Image Registration: Lucas & Kanade Revisited

L&K Derivation 1

1kh

Rx k

k

x

k hxI

hxIxIxw

xwh

)(

)]()()[(

)(

1'0

0

I0(x+hk)

I(x)

Page 22: Iterative Image Registration: Lucas & Kanade Revisited

L&K Derivation 1

1kh

Rx k

k

x

k hxI

hxIxIxw

xwh

)(

)]()()[(

)(

1'0

0

I0(x+hf)

I(x)

Page 23: Iterative Image Registration: Lucas & Kanade Revisited

Lucas & KanadeDerivation

#2

Page 24: Iterative Image Registration: Lucas & Kanade Revisited

L&K Derivation 2

• Sum-of-squared-difference (SSD) error

E(h) = [ I(x) - I0(x+h) ]2x R

E(h) [ I(x) - I0(x) - hI0’(x) ]2x R

Page 25: Iterative Image Registration: Lucas & Kanade Revisited

L&K Derivation 2

2[I0’(x)(I(x) - I0(x) ) - hI0’(x)2] x Rh

E

I0’(x)(I(x) - I0(x))x R h I0’(x)2

x R

= 0

Page 26: Iterative Image Registration: Lucas & Kanade Revisited

Comparison

I0’(x)[I(x) - I0(x)] h I0’(x)2

x

x

h

w(x)[I(x) - I0(x)]

w(x)x

x I0’(x)

Page 27: Iterative Image Registration: Lucas & Kanade Revisited

Comparison

I0’(x)[I(x) - I0(x)] h I0’(x)2

x

h

x

w(x)[I(x) - I0(x)]

w(x)x

x I0’(x)

Page 28: Iterative Image Registration: Lucas & Kanade Revisited

Generalizations

Page 29: Iterative Image Registration: Lucas & Kanade Revisited

Original

h ) = x R

(E [I( x ) - (x ]2)+ h I

Page 30: Iterative Image Registration: Lucas & Kanade Revisited

Original

• Dimension of image

h ) = x R

(E [I( x ) - (x ]2)+ h

1-dimensional

I

LK BAHH ST S BJ HB BL G SI CETSC

Page 31: Iterative Image Registration: Lucas & Kanade Revisited

Generalization 1a

• Dimension of image

h ) = x R

(E [I( x ) - (x ]2)+ h

y

xx2D:

I

LK BAHH ST S BJ HB BL G SI CETSC

Page 32: Iterative Image Registration: Lucas & Kanade Revisited

Generalization 1b

• Dimension of image

h ) = x R

(E [I( x ) - (x ]2)+ h

1

y

x

xHomogeneous 2D:

I

LK BAHH ST S BJ HB BL G SI CETSC

Page 33: Iterative Image Registration: Lucas & Kanade Revisited

Problem A

LK BAHH ST S BJ HB BL G SI CETSC

Does the iteration converge?

Page 34: Iterative Image Registration: Lucas & Kanade Revisited

Problem A

Local minima:

Page 35: Iterative Image Registration: Lucas & Kanade Revisited

Problem A

Local minima:

Page 36: Iterative Image Registration: Lucas & Kanade Revisited

Problem B

- I0’(x)(I(x) - I0(x))x R h I0’(x)2

x R

h is undefined if I0’(x)2 is zerox R

LK BAHH ST S BJ HB BL G SI CETSC

Zero gradient:

Page 37: Iterative Image Registration: Lucas & Kanade Revisited

Problem B

Zero gradient:

?

Page 38: Iterative Image Registration: Lucas & Kanade Revisited

Problem B’

- (x)(I(x) - I0(x))x R

hy 2

x R

y

I )(0 xy

I

)(0 x

Aperture problem:

LK BAHH ST S BJ HB BL G SI CETSC

Page 39: Iterative Image Registration: Lucas & Kanade Revisited

Problem B’

No gradient along one direction:

?

Page 40: Iterative Image Registration: Lucas & Kanade Revisited

Solutions to A & B

• Possible solutions:– Manual intervention

LK BAHH ST S BJ HB BL G SI CETSC

Page 41: Iterative Image Registration: Lucas & Kanade Revisited

• Possible solutions:– Manual intervention– Zero motion default

LK BAHH ST S BJ HB BL G SI CETSC

Solutions to A & B

Page 42: Iterative Image Registration: Lucas & Kanade Revisited

• Possible solutions:– Manual intervention– Zero motion default– Coefficient “dampening”

LK BAHH ST S BJ HB BL G SI CETSC

Solutions to A & B

Page 43: Iterative Image Registration: Lucas & Kanade Revisited

• Possible solutions:– Manual intervention– Zero motion default– Coefficient “dampening”– Reliance on good features

LK BAHH ST S BJ HB BL G SI CETSC

Solutions to A & B

Page 44: Iterative Image Registration: Lucas & Kanade Revisited

• Possible solutions:– Manual intervention– Zero motion default– Coefficient “dampening”– Reliance on good features– Temporal filtering

LK BAHH ST S BJ HB BL G SI CETSC

Solutions to A & B

Page 45: Iterative Image Registration: Lucas & Kanade Revisited

• Possible solutions:– Manual intervention– Zero motion default– Coefficient “dampening”– Reliance on good features– Temporal filtering– Spatial interpolation / hierarchical estimation

LK BAHH ST S BJ HB BL G SI CETSC

Solutions to A & B

Page 46: Iterative Image Registration: Lucas & Kanade Revisited

• Possible solutions:– Manual intervention– Zero motion default– Coefficient “dampening”– Reliance on good features– Temporal filtering– Spatial interpolation / hierarchical estimation– Higher-order terms

LK BAHH ST S BJ HB BL G SI CETSC

Solutions to A & B

Page 47: Iterative Image Registration: Lucas & Kanade Revisited

Original

h ) = x R

(E [I( x ) - (x ]2)+ h I

Page 48: Iterative Image Registration: Lucas & Kanade Revisited

Original

• Transformations/warping of image

h ) = x R

(E [I( x ) -I(x ]2)+ h

Translations:

y

x

h

LK BAHH ST S BJ HB BL G SI CETSC

Page 49: Iterative Image Registration: Lucas & Kanade Revisited

Problem C

What about other types of motion?

Page 50: Iterative Image Registration: Lucas & Kanade Revisited

Generalization 2a

• Transformations/warping of image

A, h) = x R

(E [I(Ax ) - (x ]2)+h

Affine:

dc

baA

y

x

h

I

LK BAHH ST S BJ HB BL G SI CETSC

Page 51: Iterative Image Registration: Lucas & Kanade Revisited

Generalization 2a

Affine:

dc

baA

y

x

h

Page 52: Iterative Image Registration: Lucas & Kanade Revisited

Generalization 2b

• Transformations/warping of image

A ) = x R

(E [I( A x ) - (x ]2)

Planar perspective:

187

654

321

aa

aaa

aaa

A

I

LK BAHH ST S BJ HB BL G SI CETSC

Page 53: Iterative Image Registration: Lucas & Kanade Revisited

Generalization 2b

Planar perspective:

187

654

321

aa

aaa

aaa

A

Affine +

Page 54: Iterative Image Registration: Lucas & Kanade Revisited

Generalization 2c

• Transformations/warping of image

h ) = x R

(E [I( f(x, h) ) - (x ]2)

Other parametrized transformations

I

LK BAHH ST S BJ HB BL G SI CETSC

Page 55: Iterative Image Registration: Lucas & Kanade Revisited

Generalization 2c

Other parametrized transformations

Page 56: Iterative Image Registration: Lucas & Kanade Revisited

Problem B”

-(JTJ)-1 J (I(f(x,h)) - I0(x)) h ~

Generalized aperture problem:

LK BAHH ST S BJ HB BL G SI CETSC

- I0’(x)(I(x) - I0(x))x R h I0’(x)2

x R

Page 57: Iterative Image Registration: Lucas & Kanade Revisited

Problem B”

?

Generalizedaperture problem:

Page 58: Iterative Image Registration: Lucas & Kanade Revisited

Original

h ) = x R

(E [I( x ) - (x ]2)+ h I

Page 59: Iterative Image Registration: Lucas & Kanade Revisited

Original

• Image type

h ) = x R

(E [I( x ) - (x ]2)+ h

Grayscale images

I

LK BAHH ST S BJ HB BL G SI CETSC

Page 60: Iterative Image Registration: Lucas & Kanade Revisited

Generalization 3

• Image type

h ) = x R

(E ||I( x ) -I(x ||2)+ h

Color images

LK BAHH ST S BJ HB BL G SI CETSC

Page 61: Iterative Image Registration: Lucas & Kanade Revisited

Original

h ) = x R

(E [I( x ) - (x ]2)+ h I

Page 62: Iterative Image Registration: Lucas & Kanade Revisited

Original

• Constancy assumption

h ) = x R

(E [I( x ) -I(x ]2)+ h

Brightness constancy

LK BAHH ST S BJ HB BL G SI CETSC

Page 63: Iterative Image Registration: Lucas & Kanade Revisited

Problem C

What if illumination changes?

Page 64: Iterative Image Registration: Lucas & Kanade Revisited

Generalization 4a

• Constancy assumption

h, )=x R

(E [I( x ) - I(x ]2)++ h

Linear brightness constancy

LK BAHH ST S BJ HB BL G SI CETSC

Page 65: Iterative Image Registration: Lucas & Kanade Revisited

Generalization 4a

Page 66: Iterative Image Registration: Lucas & Kanade Revisited

Generalization 4b

• Constancy assumption

h,) = x R

(E [I( x ) - B(x]2)+ h

Illumination subspace constancy

LK BAHH ST S BJ HB BL G SI CETSC

Page 67: Iterative Image Registration: Lucas & Kanade Revisited

Problem C’

What if the texture changes?

Page 68: Iterative Image Registration: Lucas & Kanade Revisited

Generalization 4c

• Constancy assumption

h,) = x R

(E [I( x ) - ]2+ h

Texture subspace constancy

B(x)

LK BAHH ST S BJ HB BL G SI CETSC

Page 69: Iterative Image Registration: Lucas & Kanade Revisited

Problem D

Convergence is slower as #parameters increases.

Page 70: Iterative Image Registration: Lucas & Kanade Revisited

• Faster convergence:– Coarse-to-fine, filtering, interpolation, etc.

LK BAHH ST S BJ HB BL G SI CETSC

Solutions to D

Page 71: Iterative Image Registration: Lucas & Kanade Revisited

• Faster convergence:– Coarse-to-fine, filtering, interpolation, etc. – Selective parametrization

Solutions to D

LK BAHH ST S BJ HB BL G SI CETSC

Page 72: Iterative Image Registration: Lucas & Kanade Revisited

• Faster convergence:– Coarse-to-fine, filtering, interpolation, etc. – Selective parametrization – Offline precomputation

Solutions to D

LK BAHH ST S BJ HB BL G SI CETSC

Page 73: Iterative Image Registration: Lucas & Kanade Revisited

• Faster convergence:– Coarse-to-fine, filtering, interpolation, etc. – Selective parametrization – Offline precomputation

• Difference decomposition

LK BAHH ST S BJ HB G SI CETSC

Solutions to D

BL

Page 74: Iterative Image Registration: Lucas & Kanade Revisited

Solutions to D

• Difference decomposition

Page 75: Iterative Image Registration: Lucas & Kanade Revisited

Solutions to D

• Difference decomposition

Page 76: Iterative Image Registration: Lucas & Kanade Revisited

• Faster convergence:– Coarse-to-fine, filtering, interpolation, etc. – Selective parametrization – Offline precomputation

• Difference decomposition

– Improvements in gradient descent

LK BAHH ST S BJ HB G SI CETSC

Solutions to D

BL

Page 77: Iterative Image Registration: Lucas & Kanade Revisited

• Faster convergence:– Coarse-to-fine, filtering, interpolation, etc. – Selective parametrization– Offline precomputation

• Difference decomposition

– Improvements in gradient descent• Multiple estimates of spatial derivatives

LK BAHH ST S BJ HB G SI CETSC

Solutions to D

BL

Page 78: Iterative Image Registration: Lucas & Kanade Revisited

Solutions to D

• Multiple estimates / state-space sampling

Page 79: Iterative Image Registration: Lucas & Kanade Revisited

Generalizations

x R

[I( x ) - (x ]2)+ h I

Modifications made so far:

Page 80: Iterative Image Registration: Lucas & Kanade Revisited

Original

• Error norm

h ) = x R

(E [I( x ) -I(x ]2)+ h

Squared difference:

LK BAHH ST S BJ HB BL G SI CETSC

Page 81: Iterative Image Registration: Lucas & Kanade Revisited

Problem E

What about outliers?

Page 82: Iterative Image Registration: Lucas & Kanade Revisited

Generalization 5a

• Error norm

h ) = x R

(E (I( x ) -I(x ))+ h

Robust error norm:

22

2

)(uk

uuρ

LK BAHH ST S BJ HB BL G SI CETSC

Page 83: Iterative Image Registration: Lucas & Kanade Revisited

Original

h ) = x R

(E [I( x ) - (x ]2)+ h I

Page 84: Iterative Image Registration: Lucas & Kanade Revisited

Original

• Image region / pixel weighting

h ) = x R

(E [I( x ) -I(x ]2)+ h

Rectangular:

LK BAHH ST S BJ HB BL G SI CETSC

Page 85: Iterative Image Registration: Lucas & Kanade Revisited

Problem E’

What about background clutter?

Page 86: Iterative Image Registration: Lucas & Kanade Revisited

Generalization 6a

• Image region / pixel weighting

h ) = x R

(E [I( x ) -I(x ]2)+ h

Irregular:

LK BAHH ST S BJ HB BL G SI CETSC

Page 87: Iterative Image Registration: Lucas & Kanade Revisited

Problem E”

What about foreground occlusion?

Page 88: Iterative Image Registration: Lucas & Kanade Revisited

Generalization 6b

• Image region / pixel weighting

h ) = x R

(E [I( x ) -I(x ]2)+ h

Weighted sum:

w(x)

LK BAHH ST S BJ HB BL G SI CETSC

Page 89: Iterative Image Registration: Lucas & Kanade Revisited

Generalizations

x R

[I( x ) - (x ]2)+ h I

Modifications made so far:

Page 90: Iterative Image Registration: Lucas & Kanade Revisited

Generalizations: Summary

= x R

(I( ) - w(x) (x ))h )(E f(x, h)

h ) = x R

(E [I( x ) - (x ]2)+ h I

Page 91: Iterative Image Registration: Lucas & Kanade Revisited

Foresight

• Lucas & Kanade (IUW 1981)

• Bergen, Anandan, Hanna, Hingorani (ECCV 1992)

• Shi & Tomasi (CVPR 1994)

• Szeliski & Coughlan (CVPR 1994)

• Szeliski (WACV 1994)

• Black & Jepson (ECCV 1996)

• Hager & Belhumeur (CVPR 1996)

• Bainbridge-Smith & Lane (IVC 1997)

• Gleicher (CVPR 1997)

• Sclaroff & Isidoro (ICCV 1998)

• Cootes, Edwards, & Taylor (ECCV 1998)

LK BAHH ST S BJ HB BL G SI CETSC

Page 92: Iterative Image Registration: Lucas & Kanade Revisited

Summary

• Generalizations– Dimension of image– Image transformations / motion models– Pixel type– Constancy assumption– Error norm– Image mask

L&K ?Y

Y

n

Y

n

Y

Page 93: Iterative Image Registration: Lucas & Kanade Revisited

Summary

• Common problems:– Local minima– Aperture effect– Illumination changes– Convergence issues– Outliers and occlusions

L&K ?Y

maybe

Y

Y

n

Page 94: Iterative Image Registration: Lucas & Kanade Revisited

• Mitigation of aperture effect:– Manual intervention– Zero motion default– Coefficient “dampening”– Elimination of poor textures– Temporal filtering– Spatial interpolation / hierarchical – Higher-order terms

Summary

L&K ?n

n

n

n

Y

Y

n

Page 95: Iterative Image Registration: Lucas & Kanade Revisited

Summary

• Better convergence:– Coarse-to-fine, filtering, etc.– Selective parametrization – Offline precomputation

• Difference decomposition

– Improvements in gradient descent• Multiple estimates of spatial derivatives

L&K ?Y

nmaybe

maybe

maybe

maybe

Page 96: Iterative Image Registration: Lucas & Kanade Revisited

Hindsight

• Lucas & Kanade (IUW 1981)

• Bergen, Anandan, Hanna, Hingorani (ECCV 1992)

• Shi & Tomasi (CVPR 1994)

• Szeliski & Coughlan (CVPR 1994)

• Szeliski (WACV 1994)

• Black & Jepson (ECCV 1996)

• Hager & Belhumeur (CVPR 1996)

• Bainbridge-Smith & Lane (IVC 1997)

• Gleicher (CVPR 1997)

• Sclaroff & Isidoro (ICCV 1998)

• Cootes, Edwards, & Taylor (ECCV 1998)


Top Related