Motion Estimation I What affects the induced image motion? Camera motion Object motion Scene...

32
Motion Estimation I What affects the induced image motion? • Camera motion • Object motion • Scene structure

Transcript of Motion Estimation I What affects the induced image motion? Camera motion Object motion Scene...

Page 1: Motion Estimation I What affects the induced image motion? Camera motion Object motion Scene structure.

Motion Estimation IWhat affects the induced image motion?

• Camera motion• Object motion• Scene structure

Page 2: Motion Estimation I What affects the induced image motion? Camera motion Object motion Scene structure.

Example Flow Fields

• This lesson – estimation of general flow-fields• Next lesson – constrained by global transformations

Page 3: Motion Estimation I What affects the induced image motion? Camera motion Object motion Scene structure.

The Aperture Problem

So how much information is there locally…?

Page 4: Motion Estimation I What affects the induced image motion? Camera motion Object motion Scene structure.

The Aperture Problem

Copyright, 1996 © Dale Carnegie & Associates, Inc.

Not enough info in local regions

Page 5: Motion Estimation I What affects the induced image motion? Camera motion Object motion Scene structure.

The Aperture Problem

Copyright, 1996 © Dale Carnegie & Associates, Inc.

Not enough info in local regions

Page 6: Motion Estimation I What affects the induced image motion? Camera motion Object motion Scene structure.

The Aperture Problem

Copyright, 1996 © Dale Carnegie & Associates, Inc.

Page 7: Motion Estimation I What affects the induced image motion? Camera motion Object motion Scene structure.

The Aperture Problem

Copyright, 1996 © Dale Carnegie & Associates, Inc.

Information is propagated from regions with high certainty (e.g., corners) to regions with low certainty.

Page 8: Motion Estimation I What affects the induced image motion? Camera motion Object motion Scene structure.

Such info propagation can cause optical illusions…

Illusory corners

Page 9: Motion Estimation I What affects the induced image motion? Camera motion Object motion Scene structure.

1. Gradient based methods (Horn &Schunk, Lucase & Kanade)

2. Region based methods (Correlation, SSD, Normalized correlation)

• Direct (intensity-based) Methods

• Feature-based Methods

Page 10: Motion Estimation I What affects the induced image motion? Camera motion Object motion Scene structure.

Image J (taken at time t)

yx,

Brightness Constancy Assumption

yxJ , vyuxI ,

Image I(taken at time t+1)

vyux ,

Page 11: Motion Estimation I What affects the induced image motion? Camera motion Object motion Scene structure.

Brightness Constancy Equation:

The Brightness Constancy Constraint

),(),( ),(),( yxyx vyuxIyxJ

),(),(),(),(),(),( yxvyxIyxuyxIyxIyxJ yx Linearizing (assuming small (u,v)):

),(),(),(),(),(),(0 yxJyxIyxvyxIyxuyxI yx

),(),(),(),(),( yxIyxvyxIyxuyxI tyx

),(),(

),(),( yxI

yxv

yxuyxI t

T

Page 12: Motion Estimation I What affects the induced image motion? Camera motion Object motion Scene structure.

* One equation, 2 unknowns

* A line constraint in (u,v) space.

* Can recover Normal Flow.

Observations:

Need additional constraints…

Page 13: Motion Estimation I What affects the induced image motion? Camera motion Object motion Scene structure.

Horn and Schunk (1981)

Add global smoothness term

2222

),(yxyx

yx

vvuu

Smoothness error

2

),(tyx

yx

IvIuIE

Error in brightness constancy equation

sc EE Minimize:

Solve by using calculus of variations

Page 14: Motion Estimation I What affects the induced image motion? Camera motion Object motion Scene structure.

Horn and Schunk (1981)

Problems…

* Smoothness assumption wrong at motion/depth discontinuities over-smoothing of the flow field.

* How is Lambda determined…?

Page 15: Motion Estimation I What affects the induced image motion? Camera motion Object motion Scene structure.

Lucas-Kanade (1984)

Windowyx

tyx IvyxIuyxIvuE),(

2),(),(),(

Assume a single displacement (u,v) for all pixels within a small window

Minimize E(u,v):

Geometrically -- Intersection of multiple line constraints

Algebraically --

Page 16: Motion Estimation I What affects the induced image motion? Camera motion Object motion Scene structure.

Lucas-Kanade (1984)

Windowyx

tyx IvyxIuyxIvuE),(

2),(),(),(

Differentiating w.r.t u and v and equating to 0:

ty

tx

yyx

yxx

II

II

v

u

III

III2

2

tT IIUII

Solve for (u,v)[ Repeat this process for each and every pixel in the image ]

Minimize E(u,v):

Page 17: Motion Estimation I What affects the induced image motion? Camera motion Object motion Scene structure.

Problems…

* Still smoothes motion discontinuities (but unlike Horn & Schunk, does not propagate error across the entire image)

* Singularities (partially solved by coarse-to-fine)

Lucas-Kanade (1984)

Page 18: Motion Estimation I What affects the induced image motion? Camera motion Object motion Scene structure.

Singularites

2

2

yyx

yxx

III

III

Where in the image will this matrix be invertible and where not…?

Homework

Page 19: Motion Estimation I What affects the induced image motion? Camera motion Object motion Scene structure.

Linearization approximation iterate & warp

xx0

Initial guess:

Estimate:

estimate update

Page 20: Motion Estimation I What affects the induced image motion? Camera motion Object motion Scene structure.

xx0

estimate update

Initial guess:

Estimate:

Linearization approximation iterate & warp

Page 21: Motion Estimation I What affects the induced image motion? Camera motion Object motion Scene structure.

xx0

Initial guess:

Estimate:

Initial guess:

Estimate:

estimate update

Linearization approximation iterate & warp

Page 22: Motion Estimation I What affects the induced image motion? Camera motion Object motion Scene structure.

xx0

Linearization approximation iterate & warp

Page 23: Motion Estimation I What affects the induced image motion? Camera motion Object motion Scene structure.

Revisiting the small motion assumption

Is this motion small enough?Probably not—it’s much larger than one pixel (2nd order terms

dominate)How might we solve this problem?

Page 24: Motion Estimation I What affects the induced image motion? Camera motion Object motion Scene structure.

0 tyx IvIuI ==> small u and v ...

u=10 pixels

u=5 pixels

u=2.5 pixels

u=1.25 pixels

image Iimage J

u

iterate refine

u

+

Pyramid of image J Pyramid of image I

image Iimage J

Coarse-to-Fine EstimationAdvantages: (i) Larger displacements. (ii) Speedup. (iii) Information from multiple window sizes.

Page 25: Motion Estimation I What affects the induced image motion? Camera motion Object motion Scene structure.

Optical Flow Results

Page 26: Motion Estimation I What affects the induced image motion? Camera motion Object motion Scene structure.

Optical Flow Results

Page 27: Motion Estimation I What affects the induced image motion? Camera motion Object motion Scene structure.

1. Gradient based methods (Horn &Schunk, Lucase & Kanade, …)

2. Region based methods (SSD, Normalized correlation, etc.)

Copyright, 1996 © Dale Carnegie & Associates, Inc.

But… (despite coarse-to-fine estimation)

• rely on B.C.

• cannot handle very large motions (no more than 10%-15% of image width/height)

• small object moving fast…?

Page 28: Motion Estimation I What affects the induced image motion? Camera motion Object motion Scene structure.

Region-Based Methods

* Define a small area around a pixel as the region.

* Match the region against each pixel within a search area in next image.

* Use a match measure (e.g., sum of-squares difference, normalized correlation, etc.)

* Choose the maximum (or minimum) as the match.

Page 29: Motion Estimation I What affects the induced image motion? Camera motion Object motion Scene structure.

SSD Surface – Textured area

Page 30: Motion Estimation I What affects the induced image motion? Camera motion Object motion Scene structure.

SSD Surface -- Edge

Page 31: Motion Estimation I What affects the induced image motion? Camera motion Object motion Scene structure.

SSD – homogeneous area

[Anandan’89 - Use coarse-to-fine SSD of local windows to find matches. - Propagate information using directional confidence measures extracted from each local SSD surface]

Page 32: Motion Estimation I What affects the induced image motion? Camera motion Object motion Scene structure.

B.C. + Additional constraints:

Copyright, 1996 © Dale Carnegie & Associates, Inc.

• Increase aperture: [e.g., Lucas & Kanade]

Local singularities at degenerate image regions.

• Increase analysis window to large image regions => Global model constraints: (NEXT LESSON)

Numerically stable, but requires prior model selection:

• Planar (2D) world model

[e.g., Bergen-et-al:92, Irani-et-al:92+94, Black-et-al]

• 3D world model[e.g., Hanna-et-al:91+93, Stein & Shashua:97, Irani-et-al:1999]

• Spatial smoothness: [e.g., Horn & Schunk:81, Anandan:89]

Violated at depth/motion discontinuities