spectral matting

42
Spectral Matting A. Levin D. Lischinski and Y. Weiss. A Closed Form Solution to Natural Image Matting. IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), June 2006, New York A. Levin, A. Rav-Acha, D. Lischinski. Spectral Matting. Best paper award runner up. IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), Minneapolis, June 2007 A. Levin 1,2 , A. Rav-Acha 1 , D. Lischinski 1 . Spectral Matting. IEEE Trans. Pattern Analysis and Machine Intelligence, Oct 2008. 1 School of CS&Eng The Hebrew University 2 CSAIL MIT 1

Upload: erno

Post on 31-Jan-2016

90 views

Category:

Documents


1 download

DESCRIPTION

Spectral Matting. A. Levin D. Lischinski and Y. Weiss. A Closed Form Solution to Natural Image Matting. IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), June 2006, New York - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Spectral Matting

Spectral Matting

A. Levin D. Lischinski and Y. Weiss. A Closed Form Solution to Natural Image Matting. IEEE Conf. on Computer Vision and

Pattern Recognition (CVPR), June 2006, New York

A. Levin, A. Rav-Acha, D. Lischinski. Spectral Matting. Best paper award runner up. IEEE Conf. on Computer Vision and Pattern

Recognition (CVPR), Minneapolis, June 2007

A. Levin1,2, A. Rav-Acha1, D. Lischinski1. Spectral Matting. IEEE Trans. Pattern Analysis and Machine Intelligence, Oct 2008.

1School of CS&Eng The Hebrew University2CSAIL MIT

1

Page 2: Spectral Matting

Hard

segmentation compositing

Matte compositing

Source image

Hard segmentation and matting

2

Page 3: Spectral Matting

Previous approaches to segmentation and matting

Unsupervised

Input Hard output Matte output

Spectral segmentation:Spectral segmentation: Shi and Malik 97 Yu and Shi 03 Weiss 99 Ng et al 01 Zelnik and Perona 05 Tolliver and Miller 06

3

Page 4: Spectral Matting

Previous approaches to segmentation and matting

Unsupervised

Input Hard output Matte output

Supervised

0

1

July and Boykov01 Rother et al 04 Li et al 04

4

Page 5: Spectral Matting

Previous approaches to segmentation and matting

Unsupervised

Input Hard output Matte output

Supervised

0

1

Trimap interfaceTrimap interface: Bayesian Matting (Chuang et al 01) Poisson Matting (Sun et al 04) Random Walk (Grady et al 05)Scribbles interface:Scribbles interface: Wang&Cohen 05 Levin et al 06 Easy matting (Guan et al 06)

?

5

Page 6: Spectral Matting

User guided interface

TrimapScribbles Matting result

6

Page 7: Spectral Matting

Generalized compositing equation

iiiii BFI )1( 2 layers compositing

= x x+ 1 2L1L

7

Page 8: Spectral Matting

Generalized compositing equation

iiiii BFI )1( 2 layers compositing

= x x+ 1 2L1L

Ki

Kiii LLLI

iii ...2211

K layers compositing

= x x+

+ x x+3 4 4L3L

1 2 2L1L

Matting components

8

Page 9: Spectral Matting

Generalized compositing equation

1...21 K

iii

“Sparse” layers- 0/1 for most image pixels

Matting components:

Ki

Kiii LLLI

iii ...2211

K layers compositing

= x x+

+ x x+

10 ki

1

3 4

2 2L

4L3L

1L

9

Page 10: Spectral Matting

Automatically computed matting components

Input

1 2 3 4

8765

Unsupervised matting

10

Page 11: Spectral Matting

Building foreground object by simple components addition

=+ +

11

Page 12: Spectral Matting

Spectral segmentation

22/

),(ji CC

ejiW

WDL

j

jiWiiD ),(),(

Spectral segmentation: Analyzing smallest eigenvectors of a graph Laplacian L

E.g.: Shi and Malik 97 Yu and Shi 03 Weiss 99 Ng et al 01 Maila and shi 01 Zelnik and Perona 05 Tolliver and Miller 0612

Page 13: Spectral Matting

Problem Formulation

= x x+ 1 2L1L

Assume a and b are constant in a small window

13

Page 14: Spectral Matting

Derivation of the cost function

14

Page 15: Spectral Matting

Derivation

LJ T )(

15

Page 16: Spectral Matting

The matting Laplacian

LJ T )(

• semidefinite sparse matrix

• local function of the image:),( jiL

L

16

Page 17: Spectral Matting

The matting affinity

17

Page 18: Spectral Matting

The matting affinity

Color Distribution

Input

18

Page 19: Spectral Matting

Matting and spectral segmentation

Typical affinity function Matting affinity function

19

Page 20: Spectral Matting

Eigenvectors of input image

Input

Smallest eigenvectors 20

Page 21: Spectral Matting

Spectral segmentationFully separated classes: class indicator vectors belong to Laplacian nullspace

General case: class indicators approximated as linear combinations of smallest eigenvectors

Null

Binary indicating

vectors

Laplacian matrix

21

Page 22: Spectral Matting

Spectral segmentation

Fully separated classes: class indicator vectors belong to Laplacian nullspace

General case: class indicators approximated as linear combinations of smallest eigenvectors

Smallest eigenvectors- class indicators only up to linear transformation

33

RZero eigenvectors

Binary indicating

vectors

Laplacian matrix

Smallest eigenvecto

rs

Linear transformati

on

22

Page 23: Spectral Matting

From eigenvectors to matting components

linear transformat

ion

23

Page 24: Spectral Matting

From eigenvectors to matting components

Sparsity of matting components

Minimize

24

Page 25: Spectral Matting

From eigenvectors to matting components

Minimize

Newton’s method

with initialization

25

Page 26: Spectral Matting

From eigenvectors to matting components

Smallest eigenvectors

Projection into eigs space kCTk mEE

....

K-means

..

kCmle

1) Initialization: projection of hard segments

2) Non linear optimization for sparse components26

Page 27: Spectral Matting

Extracted Matting Components

27

Page 28: Spectral Matting

Brief Summary

LJ T )(

Construct Matting Laplacian

Smallest eigenvectors

Linear Transformation

Matting components

28

Page 29: Spectral Matting

Grouping Components

=+ +

29

Page 30: Spectral Matting

Grouping Components

Unsupervised matting User-guided matting

Complete foreground matte

=+ +

30

Page 31: Spectral Matting

Unsupervised matting

LJ T )(

Matting cost function

Hypothesis:Generate indicating vector b

31

Page 32: Spectral Matting

Unsupervised matting results

32

Page 33: Spectral Matting

User-guided matting Graph cut method

Energy function

Unary term Pairwise termConstrained components

33

Page 34: Spectral Matting

Components with the scribble interface

Components (our

approach)

Levin et al cvpr06

Wang&Cohen 05

Random Walk

Poisson 34

Page 35: Spectral Matting

Components with the scribble interface

Components (our

approach)

Levin et al cvpr06

Wang&Cohen 05

Random Walk

Poisson 35

Page 36: Spectral Matting

Direct component picking interface

=+ +

Building foreground object by simple components addition

36

Page 37: Spectral Matting

Results

37

Page 38: Spectral Matting

Quantitative evaluation

38

Page 39: Spectral Matting

Spectral matting versus obtaining trimaps from a hard segmentation

39

Page 40: Spectral Matting

Limitations Number of eigenvectors

Ground truth matte Matte from 70 eigenvectors

Matte from 400 eigenvectors40

Page 41: Spectral Matting

Limitations Number of matting components

41

Page 42: Spectral Matting

Conclusion Derived analogy between hard spectral

segmentation to image matting Automatically extract matting components

from eigenvectors Automate matte extraction process and

suggest new modes of user interaction

42