cse 291 seminar presentation andrew cosand ece cvrr

28
An Algorithm for Associating the Features of Two Images / G. L. Scott, H. C. Longuet-Higgins A direct method for stereo correspondence based on singular value decomposition / M. Pilu CSE 291 Seminar Presentation Andrew Cosand ECE CVRR

Upload: raanan

Post on 31-Jan-2016

26 views

Category:

Documents


0 download

DESCRIPTION

An Algorithm for Associating the Features of Two Images / G. L. Scott, H. C. Longuet-Higgins A direct method for stereo correspondence based on singular value decomposition / M. Pilu. CSE 291 Seminar Presentation Andrew Cosand ECE CVRR. Outline. Correspondence Problem Examples Discrepancy - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: CSE 291 Seminar Presentation Andrew Cosand ECE CVRR

An Algorithm for Associating the Features of Two Images / G. L. Scott, H. C. Longuet-Higgins

A direct method for stereo correspondence based

on singular value decomposition / M. Pilu

CSE 291 Seminar Presentation

Andrew Cosand

ECE CVRR

Page 2: CSE 291 Seminar Presentation Andrew Cosand ECE CVRR

Outline

Correspondence Problem– Examples– Discrepancy

S&L-H Solution– Distance Measure– Singular Value Decomposition– Relation to Kernel Trick

Pilu’s Contribution

Page 3: CSE 291 Seminar Presentation Andrew Cosand ECE CVRR

Correspondence Problem

Which features in image A correspond to features in image B?

Page 4: CSE 291 Seminar Presentation Andrew Cosand ECE CVRR

Correspondence Problem

This task is trivial for humans, but difficult for computers.

Page 5: CSE 291 Seminar Presentation Andrew Cosand ECE CVRR

Correspondence Problem

• Used for stereo image pairs & motion images.

• Feature correspondence should exhibit Similarity, Proximity and Exclusivity.

• Complexity is combinatorial with number of features to compare.

Page 6: CSE 291 Seminar Presentation Andrew Cosand ECE CVRR

Stereo Imaging

Trinocular camera captures 3 images, horizontally and vertically offset.

Page 7: CSE 291 Seminar Presentation Andrew Cosand ECE CVRR

Stereo Imaging

Feature correspondence is used to extract depth information from stereo images– Distances between cameras are known.– Distances between the same feature in different

images is determined.– Distance from cameras to actual object can be

calculated.

Page 8: CSE 291 Seminar Presentation Andrew Cosand ECE CVRR

Motion Tracking

Corresponding features are tracked through sequential images to determine object or camera motion.

Object Motion Only Compound Motion

Page 9: CSE 291 Seminar Presentation Andrew Cosand ECE CVRR

Local vs. Global

Page 10: CSE 291 Seminar Presentation Andrew Cosand ECE CVRR

Discrepancy

Small scale discrepancy constrains corresponding features to be close together.– Slow object movement, slight camera motion,

narrow baseline stereo

Large scale discrepancy allows widely separated features.– Fast object movement, large camera motion,

wide baseline stereo

Page 11: CSE 291 Seminar Presentation Andrew Cosand ECE CVRR

Ternus

Page 12: CSE 291 Seminar Presentation Andrew Cosand ECE CVRR

Ternus

Page 13: CSE 291 Seminar Presentation Andrew Cosand ECE CVRR

Ternus

Page 14: CSE 291 Seminar Presentation Andrew Cosand ECE CVRR

Achieving Good Global Correspondence

Requires relationships between points– The inner product of x,y coordinates yields a

deficient feature space. (Also location biased)– Gaussian weighted distance better captures the

spatial relationships between points (location and proximity).

– S&LH provides superior sphered (decorrelated) relationship.

– Pilu adds similarity relationship.

Page 15: CSE 291 Seminar Presentation Andrew Cosand ECE CVRR
Page 16: CSE 291 Seminar Presentation Andrew Cosand ECE CVRR

Scott & Longuet-Higgins

Define a distance metric between featuresGij=e(-r

ij2/22)

Create matrix of relationships for all possible feature pairs

G11

Gij

Page 17: CSE 291 Seminar Presentation Andrew Cosand ECE CVRR

S&LH Distance Measure

Gaussian Weighted scales distance weighting (discrepancy)– Analytic with respect to distance, coordinates– Decreases monotonically with distance– Positive Definite for identical images

Page 18: CSE 291 Seminar Presentation Andrew Cosand ECE CVRR

Positive Definite Matrices

• Comparing identical feature sets yields a symmetric positive definite matrix.

• Symmetric gets us real eigenvalues.

• Positive definite has positive eigenvalues (which means real square roots).

• G = UUT = QQT => Q = U1/2

Matrix Factors Inner Product

Real

Page 19: CSE 291 Seminar Presentation Andrew Cosand ECE CVRR

Singular Value Decomposition

SVD factors a matrix into the product of two orthogonal matrices and a diagonal matrix of singular values (eigenvalues).

G = TDU, G is m-by-n, – T is orthogonal, m-by-m

– D is diag(1, 2, … p), m-by-n, p=min{m,n}

– U is orthogonal, n-by-n

Page 20: CSE 291 Seminar Presentation Andrew Cosand ECE CVRR

Scott & Longuet-Higgins

Use Singular Value Decomposition on matrix. G = TDU

Page 21: CSE 291 Seminar Presentation Andrew Cosand ECE CVRR

Scott & Longuet-Higgins

Set diagonal elements of D to 1, ‘recover’ relationship matrix.

P = TIU = TU

Eliminating singular matrix rescales data in feature space, essentially sphereing it.

Page 22: CSE 291 Seminar Presentation Andrew Cosand ECE CVRR

Scott & Longuet-Higgins

Largest feature in row and column indicates mutual best match (correspondence)

Page 23: CSE 291 Seminar Presentation Andrew Cosand ECE CVRR

Relation to Kernel Trick

Gaussian Distance is essentially a kernel– Relates to a dot product in infinite dimensionial

space.– This gives a richer feature space with useful

relationships between features.– This is why the SVD works here.

Page 24: CSE 291 Seminar Presentation Andrew Cosand ECE CVRR

Pilu’s Improvement

• Rogue features don’t correspond to anything, complicating the process.

• S&LH only deals with proximity and exclusivity.

• Similarity constraint can eliminate rogue features, which shouldn’t be similar to anything.

Page 25: CSE 291 Seminar Presentation Andrew Cosand ECE CVRR

Pilu’s Improvement

Modify relationship metric to include gray-level correlation.

Gij = (e-(Cij – 1)2/22) e(-rij2/22)

Gij = ((Cij+1) /2) e(-rij2/22)

– Adds similarity to feature space (kernel operation).– Rogue features can be eliminated because they are not

similar to anything.

Page 26: CSE 291 Seminar Presentation Andrew Cosand ECE CVRR

Results

• Achieves globally better feature matches rather than locally good matches.

• Resistant to rogue points.

Page 27: CSE 291 Seminar Presentation Andrew Cosand ECE CVRR

Summary

• S&LH essentially maps input to a rich, high dimensional feature space using kernel trick, then uses SVD to determine matches.

• Pilu improves kernel to achieve better feature space.

• Combination works well.

Page 28: CSE 291 Seminar Presentation Andrew Cosand ECE CVRR

References

This presentation drew material from the following sources– S. Belonge, Notes on Spectral Correspondence

– M. Pilu, A direct method for stereo correspondence based on singular value decomposition

• variants

– G. L. Scott, H. C. Longuet-Higgins, An Algorithm for Associating the Features of Two Images