Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

51
Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors Michael Kazhdan Michael Kazhdan Th Th o o mas Funkhouser mas Funkhouser Szymon Rusinkiewicz Szymon Rusinkiewicz Princeton University Princeton University

description

Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors. Michael Kazhdan Th o mas Funkhouser Szymon Rusinkiewicz Princeton University. Motivation. Large databases of 3D models. Computer Graphics (Princeton 3D Search Engine). Mechanical CAD - PowerPoint PPT Presentation

Transcript of Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Page 1: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Rotation Invariant Spherical Harmonic Representation of

3D Shape DescriptorsMichael KazhdanMichael Kazhdan

ThThoomas Funkhousermas FunkhouserSzymon RusinkiewiczSzymon RusinkiewiczPrinceton UniversityPrinceton University

Page 2: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Motivation

Large databases of 3D modelsLarge databases of 3D models

Mechanical CAD(National Design Repository)

Molecular Biology(Audrey Sanderson)

Computer Graphics(Princeton 3D Search Engine)

Page 3: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Retrieval Approach

3D Model ShapeDescriptor

Model Database

Nearest Neighbor

Page 4: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Shape Unchanged by Rotation

=

Page 5: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Problem

Many shape descriptors are functions that Many shape descriptors are functions that rotate with the shaperotate with the shape

Extended Gaussian Image[Horn ’84]

Spherical Attribute Image[Ikeuchi ’95]

Shape Histogram [Ankerst ’99]

Spherical Extent Function[Vranic ’00]

Reflective Symmetry Descriptor [Kazhdan ’02]

Gaussian EDT[Funkhouser ’03]

Page 6: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Goal

Compute similarity of shape descriptors Compute similarity of shape descriptors independent of rotationindependent of rotation

- = ?

Page 7: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Brute Force Approach

-

-

-

-

min (rotation)- =

Impractical for databasesImpractical for databases

Page 8: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Normalization

Use PCA to place models into a canonical Use PCA to place models into a canonical coordinate framecoordinate frame

Covariance MatrixComputation

Principal Axis Alignment

Page 9: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Normalization

Doesn’t always workDoesn’t always work• Only second order informationOnly second order information

Page 10: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Our Approach

Eliminate rotation dependence in spherical Eliminate rotation dependence in spherical and 3D descriptorsand 3D descriptors

Shape Descriptor

EGI [Horn ’84]SAI [Ikeuchi ’95]EXT [Vranic ’00]RSD [Kazhdan ’02]EDT [Funkhouser ’03]etc.

Shape Descriptor

Page 11: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Our Approach

Eliminate rotation dependence in spherical Eliminate rotation dependence in spherical and 3D descriptorsand 3D descriptors

Rotation Invariant

Representation

Shape Descriptor

Page 12: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Outline

IntroductionIntroductionBackgroundBackgroundHarmonic RepresentationHarmonic RepresentationPropertiesPropertiesExperimental ResultsExperimental ResultsConclusion and Future WorkConclusion and Future Work

Page 13: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Key Idea

Obtain rotation invariant representation by Obtain rotation invariant representation by storing amplitude and eliminating phasestoring amplitude and eliminating phase

+ + + +…=

[Lo 1989]

[Burel 1995]

Page 14: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Fourier Descriptors

CircularFunction

Page 15: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Fourier Descriptors

+ + += + …

Cosine/Sine Decomposition

CircularFunction

Page 16: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Fourier Descriptors

=

+ + +

Constant

= + …

Frequency Decomposition

CircularFunction

Page 17: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Fourier Descriptors

=

+ + +

+

Constant 1st Order

= + …+

Frequency Decomposition

CircularFunction

Page 18: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Fourier Descriptors

=

+ + +

+ +

Constant 1st Order 2nd Order

= + …+

Frequency Decomposition

CircularFunction

Page 19: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Fourier Descriptors

=

+ + +

+ + +

Constant 1st Order 2nd Order 3rd Order

= + …

+ …

+

Frequency Decomposition

CircularFunction

Page 20: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

+ + + + …+

Fourier Descriptors

= + + +

Constant 1st Order 2nd Order 3rd Order

+ …

Frequency Decomposition

=Amplitudes invariant

to rotationCircularFunction

Page 21: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Harmonic Representation

SphericalFunction

Page 22: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Harmonic Representation

=

SphericalFunction

+ + + + …

Harmonic Decomposition

Page 23: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Harmonic Representation

=

SphericalFunction

+ + + + …

+ + + +…=

Constant 1st Order 2nd Order 3rd Order

Page 24: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Harmonic Representation

+ + + +…=

Norms Invariantto Rotation

Store “how much” Store “how much” (L(L22-norm) of the -norm) of the shape resides in shape resides in each frequencyeach frequency

Page 25: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

3D Function (Voxel Grid)

Restrict to concentric spheresRestrict to concentric spheres

Page 26: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

3D Function (Voxel Grid)

=

=

=

+

+

+ +

+

+ +

+

+

+

+

+

Compute harmonic representation of each Compute harmonic representation of each sphere independentlysphere independently

Page 27: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

3D Function (Voxel Grid)

Combine harmonic representationsCombine harmonic representations

Radius

Frequency

Page 28: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Matching

LL22-difference of harmonic representations…-difference of harmonic representations…

Harmonic Representation Harmonic Representation

-2

Page 29: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Matching

min (rotations)

--2 2

… … bounds proximity of descriptors over all bounds proximity of descriptors over all rotationsrotations

Page 30: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Outline

IntroductionIntroductionBackgroundBackgroundHarmonic RepresentationHarmonic RepresentationPropertiesPropertiesExperimental ResultsExperimental ResultsConclusion and Future WorkConclusion and Future Work

Page 31: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Advantages

The harmonic representations is:The harmonic representations is:•Rotation invariantRotation invariant•Multi-resolutionMulti-resolution•CompactCompact•DiscriminatingDiscriminating

Page 32: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Compact

……

Page 33: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Compact

……

……

Page 34: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Compact

……

……

Page 35: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Compact

……

……

Page 36: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Compact

……

……

Page 37: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Information Loss

•Intra-frequency information lossIntra-frequency information loss•Cross-frequency information lossCross-frequency information loss•Cross-radial information lossCross-radial information loss

Page 38: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Information Loss (Spherical Descriptor)

•Intra-frequency information lossIntra-frequency information loss•Cross-frequency information lossCross-frequency information loss

Page 39: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Information Loss (Spherical Descriptor)

•Intra-frequency information lossIntra-frequency information loss•Cross-frequency information lossCross-frequency information loss

+

+

22.5o90o

=

=

Page 40: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Information Loss (3D Descriptor)

•Cross-radial information lossCross-radial information loss

Page 41: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Outline

IntroductionIntroductionBackgroundBackgroundHarmonic RepresentationHarmonic RepresentationPropertiesPropertiesExperimental ResultsExperimental ResultsConclusion and Future WorkConclusion and Future Work

Page 42: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Shape DescriptorsExtended Gaussian Image

Horn 1984

Spherical Extent Function Vranic 2000

Shape Histogram Ankerst 1999

Gaussian EDT Funkhouser 2003

Page 43: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Experimental Database

Viewpoint “household” databaseViewpoint “household” database1,890 models, 85 classes1,890 models, 85 classes

153 dining chairs 25 livingroom chairs 16 beds 12 dining tables

8 chests 28 bottles 39 vases 36 end tables

Page 44: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Gaussian EDT ResultsPCA-Normalized Results

Harmonic Representation Results

Query

11 22 33 44 55

66 77 88 99 1010

11 22 33 44 55

66 77 88 99 1010

Page 45: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Gaussian EDT Results

Precision vs. RecallPrecision vs. Recall

50% 100%

100%

0%

50%

Recall

Prec

isio

n

0%

HarmonicsPCA

Page 46: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Retrieval ResultsHarmonicsPCA

HarmonicsPCA

HarmonicsPCA

HarmonicsPCA

EGI

EDTEXT

SECT

50% 100%

100%

0%

50%

RecallPr

ecis

ion

0%

50% 100%

100%

0%

50%

Recall

Prec

isio

n

0%50% 100%

100%

0%

50%

Recall

Prec

isio

n

0%

50% 100%

100%

0%

50%

Recall

Prec

isio

n

0%

•EGI: Extended Gaussian Image

•SECT: Shape Histogram (Sectors)

•EXT: Spherical Extent Function

•EDT: Gaussian Euclidean Distance Transform

Page 47: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Retrieval ResultsHarmonicsPCA

EGIHarmonicsPCA

SECT

50% 100%

100%

0%

50%

RecallPr

ecis

ion

0%

HarmonicsPCA

EDT

50% 100%

100%

0%

50%

Recall

Prec

isio

n

0%

HarmonicsPCA

EXT

50% 100%

100%

0%

50%

Recall

Prec

isio

n

0%

50% 100%

100%

0%

50%

Recall

Prec

isio

n

0%

•EGI: Extended Gaussian Image

•SECT: Shape Histogram (Sectors)

•EXT: Spherical Extent Function

•EDT: Gaussian Euclidean Distance Transform

Page 48: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Exhaustive Gaussian EDT Results

HarmonicPCA

min L2

100%

50%

0%0% 50% 100%

Recall

Prec

isio

n

Gaussian EDT -

-

-

-

min (rotation)

Page 49: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Summary and Conclusion

Provide a rotation invariant representation of Provide a rotation invariant representation of shape descriptors that:shape descriptors that:• Eliminates PCA dependenceEliminates PCA dependence• Gives better matching performanceGives better matching performance• Is more compactIs more compact• Is a multi-resolution representationIs a multi-resolution representation

Page 50: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Future Work

Managing Information LossManaging Information Loss• Obtain cross radial information for 3D descriptorsObtain cross radial information for 3D descriptors• Obtain cross frequency informationObtain cross frequency information• Get finer resolution of rotation invariance within Get finer resolution of rotation invariance within

frequenciesfrequencies

More GenerallyMore Generally• Consider new shape descriptorsConsider new shape descriptors

Page 51: Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors

Thank YouFundingFunding

National Science FoundationNational Science Foundation

Sloan FoundationSloan Foundation

Spherical HarmonicsSpherical HarmonicsDan Rockmore and Peter KostelecDan Rockmore and Peter Kostelec

SpharmonicKit:SpharmonicKit:

http://www.cs.dartmouth.edu/~geelong/spherehttp://www.cs.dartmouth.edu/~geelong/sphere3D Shape Matching3D Shape Matching

Patrick Min, Alex Halderman, Phil Shilane, David Jacobs, Joyce ChenPatrick Min, Alex Halderman, Phil Shilane, David Jacobs, Joyce Chen

Princeton 3D Model Search Engine:Princeton 3D Model Search Engine:

http://shape.cs.princeton.eduhttp://shape.cs.princeton.edu