+ All Categories
Home > Documents > Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge...

Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge...

Date post: 21-Dec-2015
Category:
View: 218 times
Download: 3 times
Share this document with a friend
Popular Tags:
46
Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C.
Transcript
Page 1: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.

Efficient Spatiotemporal Grouping Using the Nyström

Method

Charless Fowlkes, U.C. Berkeley

Serge Belongie, U.C. San Diego

Jitendra Malik, U.C. Berkeley

Page 2: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.

Grouping With Pairwise Affinities

Sarkar and Boyer (1996), Shi and Malik (1997), Perona and Freeman (1998), Gdalyahu, Weinshall, and Werman (2000), .....

• Compute the similarities between pairs of points in the image

• Find groups of points which have high similarity with each other and low similarity with the rest of the image.

Page 3: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.

Normalized Cuts

2. Find the leading eigenvectors of the Normalized Laplacian

3. Segment the image using the leading eigenvectors

1. Compute matrix K which contains the pairwise similarities

Computational Complexity: Need to find eigenvectors of anNxN matrix where N is the number of pixels.

Other spectral partitioning techniques have same complexity.

Page 4: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.

Spatiotemporal grouping

Adelson and Bergen (1985), Bolles, Baker and Marimont (1987), Shi and Malik (1998)

Page 5: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.

Computational Problem

• Hard to exploit pairwise clustering techniques since 256x384x30 frames entails ~1013 pairwise similarities.

• How can we overcome this problem?

Page 6: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.

Coping with the computational burden

1 Zero out small entries in the affinity matrix Shi and Malik (97,98)

2 Exploit redundancy between rows of the affinity matrix (this talk)

Page 7: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.

Outline

• Exploiting Redundancy• The Nyström approximation• Application to segmenting video sequences

Page 8: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.

Exploiting Redundancy

Page 9: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.

Exploiting Redundancy

Compute Affinity Matrix

Page 10: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.

Exploiting Redundancy

Compute LeadingEigenvectors of Normalized Laplacian

Page 11: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.

Exploiting Redundancy

Choose sample points

Page 12: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.

Exploiting Redundancy

Compute strip of KCompute strip of K

Page 13: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.

Exploiting Redundancy

Page 14: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.

Exploiting Redundancy

Interpolate complete eigenvectors

Page 15: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.

Outline

• Exploiting Redundancy• The Nyström approximation• Application to segmenting video sequences

Page 16: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.

Approximating eigenfunctions

E. J. Nyström (1929) Baker (1977) Williams and Seeger (2001)

Interpolate eigenfunctions using The Nyström Extension:

We would like to find numerical solutions to:

Page 17: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.

Matrix Completion

Affinity Matrix:

Approximate it with:

Approximation Error:

Page 18: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.

Approximation Error

Page 19: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.

Extrapolating Eigenvectors

Just matrix notation for the Nyström extension

Diagonalize approximate K to get complete eigenvectors

Page 20: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.

Nyström-NCuts Algorithm

1 Choose sample points in image

2 Compute similarities for A and B blocks of K

3 Compute row sums to estimate degree

4 Normalize A and B blocks by degree

5 Compute approximate eigenvectors and orthogonalize

6 Cluster the embedded points using k-means

Page 21: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.

Outline

• Exploiting Redundancy• The Nyström approximation• Application to segmenting video sequences

Page 22: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.

Affinity Function for VideoPairwise affinity function between pixels in a video sequence makes of three cues

-Similarity in color-Proximity in time and space-Common Fate (similarity in optical flow)

We use squared-exponential kernel with diagonal weighting

Page 23: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.

Performance

• Segmenting a 5 frame video sequence at 120x150 resolution (~100,000 pixels) takes less than 1 minute in MATLAB on a PC

Page 24: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.
Page 25: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.
Page 26: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.
Page 27: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.
Page 28: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.
Page 29: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.
Page 30: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.
Page 31: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.
Page 32: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.
Page 33: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.
Page 34: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.
Page 35: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.
Page 36: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.
Page 37: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.
Page 38: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.
Page 39: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.
Page 40: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.
Page 41: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.

Conclusions

• Applied the Nystrom approximation to Normalized Cuts

• Exploited redundancy in image sequences in order to perform efficient spatiotemporal grouping

Page 42: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.
Page 43: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.
Page 44: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.
Page 45: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.

K-Way Normalized Cuts

Find the leading eigenvectors of Normalized Laplacian

Embed data and cluster

jth pixel

ith vector

V

Page 46: Efficient Spatiotemporal Grouping Using the Nyström Method Charless Fowlkes, U.C. Berkeley Serge Belongie, U.C. San Diego Jitendra Malik, U.C. Berkeley.

Sampling the image


Recommended