Polarized 3D

101
MIT Media Lab 2 Polarized 3D Achuta Kadambi, MIT Media Lab Joint work with Vahe Taamazyan, Boxin Shi, and Ramesh Raskar

Transcript of Polarized 3D

Page 1: Polarized 3D

MIT Media Lab 2

Polarized 3DAchuta Kadambi, MIT Media Lab

Joint work with Vahe Taamazyan, Boxin Shi, and Ramesh Raskar

Page 2: Polarized 3D

MIT Media Lab 3

Page 3: Polarized 3D

MIT Media Lab 4

Microsoft Kinect v2

Page 4: Polarized 3D

MIT Media Lab 5

Microsoft Kinect v2

Page 5: Polarized 3D

MIT Media Lab 6

Microsoft Kinect v2

Page 6: Polarized 3D

MIT Media Lab 7

Page 7: Polarized 3D

MIT Media Lab 8

Multistripe Laser Scan

Page 8: Polarized 3D

MIT Media Lab 9

Multistripe Laser Scan

Page 9: Polarized 3D

MIT Media Lab 10

Multistripe Laser Scan

NextEngine 3D$3000 USDRaster

Page 10: Polarized 3D

MIT Media Lab 11

Multistripe Laser Scan

NextEngine 3D$3000 USDRaster

Page 11: Polarized 3D

MIT Media Lab 12

Page 12: Polarized 3D

MIT Media Lab 13

Polarized 3D

3D Photo w. $30 Pol. Filter

Page 13: Polarized 3D

MIT Media Lab 14

Polarized 3D

3D Photo w.$30 Pol. Filter

Page 14: Polarized 3D

MIT Media Lab 15

Polarized 3D

3D Photo w.$30 Pol. Filter

Page 15: Polarized 3D

MIT Media Lab 16

Plenoptic Light Transport

Adelson and Bergen “The Plenoptic Function…” MIT Press 1991

Page 16: Polarized 3D

MIT Media Lab 17

Plenoptic Light Transport

Viewpoint Diversity (Light Field Cam)

Adelson and Bergen “The Plenoptic Function…” MIT Press 1991

Page 17: Polarized 3D

MIT Media Lab 18

Plenoptic Light Transport

Viewpoint Diversity (Light Field Cam)

Wavelength Diversity (Hyperspectral Cam)

Adelson and Bergen “The Plenoptic Function…” MIT Press 1991

Page 18: Polarized 3D

MIT Media Lab 19

Plenoptic Light Transport

Viewpoint Diversity (Light Field Cam)

Wavelength Diversity (Hyperspectral Cam)

Polarization Diversity(Photos, Shape, Scatter)

Adelson and Bergen “The Plenoptic Function…” MIT Press 1991

Page 19: Polarized 3D

MIT Media Lab 20

Plenoptic Light Transport

Viewpoint Diversity (Light Field Cam)

Wavelength Diversity (Hyperspectral Cam)

Polarization Diversity(Photos, Shape, Scatter)

Time of Flight(3D, Scattering)

Adelson and Bergen “The Plenoptic Function…” MIT Press 1991

Page 20: Polarized 3D

MIT Media Lab 21

Plenoptic Light Transport

Viewpoint Diversity (Light Field Cam)

Wavelength Diversity (Hyperspectral Cam)

Polarization Diversity(Photos, Shape, Scatter)

Time of Flight(3D, Scattering)

Bounce Index(Scattering)

Adelson and Bergen “The Plenoptic Function…” MIT Press 1991

Page 21: Polarized 3D

MIT Media Lab 22

Plenoptic Light Transport

Viewpoint Diversity (Light Field Cam)

Wavelength Diversity (Hyperspectral Cam)

Polarization Diversity(Photos, Shape, Scatter)

Time of Flight(3D, Scattering)

Bounce Index(Scattering)

Adelson and Bergen “The Plenoptic Function…” MIT Press 1991

Page 22: Polarized 3D

MIT Media Lab 23

Plenoptic Light Transport

Viewpoint Diversity (Light Field Cam)

Wavelength Diversity (Hyperspectral Cam)

Polarization Diversity(Photos, Shape, Scatter)

Time of Flight(3D, Scattering)

Bounce Index(Scattering)

Adelson and Bergen “The Plenoptic Function…” MIT Press 1991

Page 23: Polarized 3D

MIT Media Lab 24

Plenoptic Light Transport

Viewpoint Diversity (Light Field Cam)

Wavelength Diversity (Hyperspectral Cam)

Polarization Diversity(Photos, Shape, Scatter)

Time of Flight(3D, Scattering)

Bounce Index(Scattering)

Adelson and Bergen “The Plenoptic Function…” MIT Press 1991

Page 24: Polarized 3D

MIT Media Lab 25

Plenoptic Light Transport

Viewpoint Diversity (Light Field Cam)

Wavelength Diversity (Hyperspectral Cam)

Polarization Diversity(Photos, Shape, Scatter)

Time of Flight(3D, Scattering)

Bounce Index(Scattering)

Adelson and Bergen “The Plenoptic Function…” MIT Press 1991

Page 25: Polarized 3D

MIT Media Lab 26

Plenoptic Light Transport

Viewpoint Diversity (Light Field Cam)

Wavelength Diversity (Hyperspectral Cam)

Polarization Diversity(Photos, Shape, Scatter)

Time of Flight(3D, Scattering)

Bounce Index(Scattering)

Adelson and Bergen “The Plenoptic Function…” MIT Press 1991

Page 26: Polarized 3D

MIT Media Lab 27

Plenoptic Light Transport

Viewpoint Diversity (Light Field Cam)

Wavelength Diversity (Hyperspectral Cam)

Polarization Diversity(Photos, Shape, Scatter)

Time of Flight(3D, Scattering)

Bounce Index(Scattering)

Adelson and Bergen “The Plenoptic Function…” MIT Press 1991

Page 27: Polarized 3D

MIT Media Lab 28

Plenoptic Light Transport

Viewpoint Diversity (Light Field Cam)

Wavelength Diversity (Hyperspectral Cam)

Polarization Diversity(Photos, Shape, Scatter)

Time of Flight(3D, Scattering)

Bounce Index(Scattering)

Adelson and Bergen “The Plenoptic Function…” MIT Press 1991

Page 28: Polarized 3D

MIT Media Lab 29

Polarization of Light

Page 29: Polarized 3D

MIT Media Lab 30

Polarization of Light

Page 30: Polarized 3D

MIT Media Lab 31

Polarization of Light

Plane of Polarization

Page 31: Polarized 3D

MIT Media Lab 32

Polarization of Light

Plane of Polarization Plane of Polarization

Page 32: Polarized 3D

MIT Media Lab 33

Brewster’s Angle

Page 33: Polarized 3D

MIT Media Lab 34

Cool 2D Photos

Photo Credit: Bob Atkins

Page 34: Polarized 3D

MIT Media Lab 35

Cool 2D Photos

Photo Credit: Bob Atkins

Page 35: Polarized 3D

MIT Media Lab 36

Polarization used in 2D photography…

… But what about Polarizers for 3D Cams?

Page 36: Polarized 3D

MIT Media Lab 37

Shape from Polarization

Page 37: Polarized 3D

MIT Media Lab 38

Shape from Polarization

Page 38: Polarized 3D

MIT Media Lab 39

Shape from Polarization

Page 39: Polarized 3D

MIT Media Lab 40

Shape from Polarization

Page 40: Polarized 3D

MIT Media Lab 41

Shape from Polarization

Page 41: Polarized 3D

MIT Media Lab 42

Shape from Polarization

Old Principle [Fresnel 1819]

cos cosc oso cs

i

i t

tnrn

θ θθ θ⊥ +−

=

coscos cco

oss i t

t i

rnnθ θ

θ θ+−

=

SfP crux: Solve for theta

Page 42: Polarized 3D

MIT Media Lab 43

Shape from Polarization

Old Principle [Fresnel 1819]

cos cosc oso cs

i

i t

tnrn

θ θθ θ⊥ +−

=

coscos cco

oss i t

t i

rnnθ θ

θ θ+−

=

SfP crux: Solve for theta

Need to know refractive index

Page 43: Polarized 3D

MIT Media Lab 44

Page 44: Polarized 3D

MIT Media Lab 45

Page 45: Polarized 3D

MIT Media Lab 46

Page 46: Polarized 3D

MIT Media Lab 47

Page 47: Polarized 3D

MIT Media Lab 48

Page 48: Polarized 3D

MIT Media Lab 49

Page 49: Polarized 3D

MIT Media Lab 50

Page 50: Polarized 3D

MIT Media Lab 51

Page 51: Polarized 3D

MIT Media Lab 52

Page 52: Polarized 3D

MIT Media Lab 53

Page 53: Polarized 3D

MIT Media Lab 54

Can use Schechner 15 ICCP

Page 54: Polarized 3D

MIT Media Lab 55

Image Formation Model

Page 55: Polarized 3D

MIT Media Lab 56

Image Formation Model

( )( )max min max minpol pol( ) cos 2

2 2I II II φ φ φ+

+ −= −

Page 56: Polarized 3D

MIT Media Lab 57

Image Formation Model

( )( )max min max minpol pol( ) cos 2

2 2I II II φ φ φ+

+ −= −

Suppose and 'φ φ φ π∃ = +

Page 57: Polarized 3D

MIT Media Lab 58

Image Formation Model

( )( )max min max minpol pol( ) cos 2

2 2I II II φ φ φ+

+ −= −

Suppose and 'φ φ φ π∃ = +

Azimuthal Ambiguity problem with 2 solutionsP

Page 58: Polarized 3D

MIT Media Lab 59

Why is Shape from Polarization Unpopular?

1. \pi Ambiguity in Surface Normal

Page 59: Polarized 3D

MIT Media Lab 60

Why is Shape from Polarization Unpopular?

1. \pi Ambiguity in Surface Normal

Page 60: Polarized 3D

MIT Media Lab 61

Why is Shape from Polarization Unpopular?

1. \pi Ambiguity in Surface Normal

2. Refractive Distortion

Page 61: Polarized 3D

MIT Media Lab 62

Why is Shape from Polarization Unpopular?

1. \pi Ambiguity in Surface Normal

2. Refractive Distortion

Page 62: Polarized 3D

MIT Media Lab 63

Why is Shape from Polarization Unpopular?

1. \pi Ambiguity in Surface Normal

2. Refractive Distortion

3. Low SNR for some geometries

Page 63: Polarized 3D

MIT Media Lab 64

Why is Shape from Polarization Unpopular?

1. \pi Ambiguity in Surface Normal

2. Refractive Distortion

3. Low SNR for some geometries

4. Usual challenges of integrating surface normals..

Page 64: Polarized 3D

MIT Media Lab 65

Shape from Polarization

Miyazaki ICCV 2003 Atkinson TIP 2006

Polarization Inverse Rendering Shape from Diffuse Polarization

Page 65: Polarized 3D

MIT Media Lab 66

Shape from Polarization

Miyazaki ICCV 2003 Atkinson TIP 2006

Polarization Inverse Rendering Shape from Diffuse Polarization

SfP never as popular as shading or photometric stereo

Page 66: Polarized 3D

MIT Media Lab 67

Potential of Polarization?

• Passive capture

• Break past Lambertian assumption

• No baseline

• Robust to Lighting

Page 67: Polarized 3D

MIT Media Lab 68

Frequency Analysis

Page 68: Polarized 3D

MIT Media Lab 69

Frequency Analysis

Page 69: Polarized 3D

MIT Media Lab 70

Frequency Analysis

Page 70: Polarized 3D

MIT Media Lab 71

Frequency Analysis

Page 71: Polarized 3D

MIT Media Lab 72

Frequency Analysis

Page 72: Polarized 3D

MIT Media Lab 73

Frequency Analysis

Page 73: Polarized 3D

MIT Media Lab 74

Frequency Analysis

Page 74: Polarized 3D

MIT Media Lab 75

Polarized 3D Fuses Depth and Polarization

Spanning Tree Integration

Page 75: Polarized 3D

MIT Media Lab 76

Polarized 3D Fuses Depth and Polarization

Spanning Tree Integration

Page 76: Polarized 3D

MIT Media Lab 77

Polarized 3D Fuses Depth and Polarization

Spanning Tree Integration

Page 77: Polarized 3D

MIT Media Lab 78

Polarized 3D Fuses Depth and Polarization

Spanning Tree Integration

Page 78: Polarized 3D

MIT Media Lab 79

Polarized 3D Fuses Depth and Polarization

Spanning Tree Integration

Page 79: Polarized 3D

MIT Media Lab 80

Visual Debugging

Kinect 3 Polar Photos

Page 80: Polarized 3D

MIT Media Lab 81

Visual Debugging

Kinect 3 Polar Photos SfP

Page 81: Polarized 3D

MIT Media Lab 82

Visual Debugging

Kinect 3 Polar Photos SfP Grad. Corr.

Page 82: Polarized 3D

MIT Media Lab 83

Visual Debugging

Kinect 3 Polar Photos SfP Grad. Corr. Spanning Tree

Page 83: Polarized 3D

MIT Media Lab 84

Versus Expensive Laser Scanner?

Polarized 3D NextEngineMultistripe

Page 84: Polarized 3D

MIT Media Lab 85

Assumptions

Unpolarized World Assumption

Dielectric or Low-frequency Material Transition

No specular interreflections

Diffuse-dominant or Specular-dominant surfaces with slack

Page 85: Polarized 3D

MIT Media Lab 86

Break the Lambertian Assumption

Page 86: Polarized 3D

MIT Media Lab 87

Break the Lambertian AssumptionKinect

Page 87: Polarized 3D

MIT Media Lab 88

Break the Lambertian AssumptionShading [Wu 14]Kinect

Page 88: Polarized 3D

MIT Media Lab 89

Break the Lambertian AssumptionPolarized 3DShading [Wu 14]Kinect

Page 89: Polarized 3D

MIT Media Lab 90

Break the Lambertian AssumptionPolarized 3DShading [Wu 14]Kinect

Page 90: Polarized 3D

MIT Media Lab 91

Break Lighting Assumptions

Kinect

Page 91: Polarized 3D

MIT Media Lab 92

Break Lighting Assumptions

Polarized 3DShading [Wu 14]Kinect

Page 92: Polarized 3D

MIT Media Lab 93

Break Lighting Assumptions

Polarized 3DShading [Wu 14]Kinect

Page 93: Polarized 3D

MIT Media Lab 94

Break Lighting Assumptions

Polarized 3DShading [Wu 14]Kinect

Page 94: Polarized 3D

MIT Media Lab 95

Polarized 3D

Collaborators:Vage Taamazyan

Boxin ShiRamesh Raskar

DIY Guide planned for 2016

Thanks to:Gary Atkinson

Terry BoultIn Kyu Park

media.mit.edu/~achoo/[email protected]

Page 95: Polarized 3D

MIT Media Lab 96

Page 96: Polarized 3D

MIT Media Lab 97

Page 97: Polarized 3D

MIT Media Lab 98

Page 98: Polarized 3D

MIT Media Lab 99

Page 99: Polarized 3D

MIT Media Lab 100

Polarized 3DAchuta Kadambi, MIT Media Lab

Joint work with Vahe Taamazyan, Boxin Shi, and Ramesh Raskar

Page 100: Polarized 3D

MIT Media Lab 101

Polarized 3DAchuta Kadambi, MIT Media Lab

Joint work with Vahe Taamazyan, Boxin Shi, and Ramesh Raskar

Page 101: Polarized 3D

MIT Media Lab 102

Polarized 3D Power of Physics VisionAchuta Kadambi, MIT Media Lab

Joint work with Vahe Taamazyan, Boxin Shi, and Ramesh Raskar