+ All Categories
Home > Documents > IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1...

IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1...

Date post: 01-Jan-2016
Category:
Upload: rachel-tucker
View: 220 times
Download: 1 times
Share this document with a friend
Popular Tags:
36
NEAR REAL-TIME STEREO MATCHING USING GEODESIC DIFFUSION IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1 Leonardo De-Maeztu, Arantxa Villanueva, Member, IEEE, and Rafael Cabeza Guan-Yu Liu
Transcript
Page 1: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1 Leonardo De-Maeztu, Arantxa Villanueva, Member, IEEE, and.

1

NEAR REAL-TIME STEREO MATCHING USING GEODESIC DIFFUSION

IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY

2012

Leonardo De-Maeztu,Arantxa Villanueva, Member, IEEE, and

Rafael Cabeza

Guan-Yu Liu

Page 2: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1 Leonardo De-Maeztu, Arantxa Villanueva, Member, IEEE, and.

2

Outline Introduction Overview Related work Method Experimental Results CUDA Q & A

Page 3: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1 Leonardo De-Maeztu, Arantxa Villanueva, Member, IEEE, and.

3

Introduction(1/4)

Stereo matching Local matching

A finite region(window size) is being computed

Global matching Do smoothness by energy minimization techniques

Page 4: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1 Leonardo De-Maeztu, Arantxa Villanueva, Member, IEEE, and.

4

Introduction(2/4)

When using local support regions, it is implicitly assumed that all pixels in the region are of the same depth. the fronto-parallel surfaces assumption Adaptive-weight methods

Page 5: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1 Leonardo De-Maeztu, Arantxa Villanueva, Member, IEEE, and.

5

Introduction(3/4)

Adaptive-weight methods are the local algorithms yielding the best results. Highly time-consuming task

Anisotropic diffusion, a computer vision technique very similar to adaptive weighting but computationally less expensive. a computer vision technique very similar to adaptive

weighting but computationally less expensive.

Page 6: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1 Leonardo De-Maeztu, Arantxa Villanueva, Member, IEEE, and.

6

Introduction(4/4)

Geodesic diffusion is inspired by anisotropic diffusion. diffusing both matching costs and weights.

Near real-time execution is demonstrated using a commercial graphics card.

Page 7: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1 Leonardo De-Maeztu, Arantxa Villanueva, Member, IEEE, and.

7

Related Work

Adaptive-weight methods [7] Adaptive-weight methods [8] Anisotropic diffusion [9]

[7] K.-J. Yoon and I.S. Kweon, “Adaptive Support-Weight Approach forCorrespondence Search,” IEEE Trans. Pattern Analysis and MachineIntelligence, vol. 28, no. 4, pp. 650-656, Apr. 2006.

[8] A. Hosni, M. Bleyer, M. Gelautz, and C. Rhemann, “Local Stereo MatchingUsing Geodesic Support Weights,” Proc. Int’l Conf. Image Processing,pp. 2093-2096, 2009.

[9] P. Perona and J. Malik, “Scale-Space and Edge Detection Using AnisotropicDiffusion,” IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 12,no. 7, pp. 629-639, July 1990.

Page 8: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1 Leonardo De-Maeztu, Arantxa Villanueva, Member, IEEE, and.

8

Related Work(1/6)

Adaptive-weight methods [7]

Page 9: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1 Leonardo De-Maeztu, Arantxa Villanueva, Member, IEEE, and.

9

Related Work(2/6)

Adaptive-weight methods [7]

123

123

123

truncated absolute difference (TAD)

Euclidean distance between the values in the CIELab color space and spatial euclidean distance

Page 10: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1 Leonardo De-Maeztu, Arantxa Villanueva, Member, IEEE, and.

10

Related Work(3/6)

Adaptive-weight methods [8]

Page 11: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1 Leonardo De-Maeztu, Arantxa Villanueva, Member, IEEE, and.

11

Related Work(4/6)

Adaptive-weight methods [8]

123

123

123

Shortest path

Page 12: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1 Leonardo De-Maeztu, Arantxa Villanueva, Member, IEEE, and.

12

Related Work(5/6)

The two algorithms use the same optimization technique, winner-takes-all (WTA).

Page 13: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1 Leonardo De-Maeztu, Arantxa Villanueva, Member, IEEE, and.

13

Related Work(6/6)

Anisotropic diffusion is a computer vision technique similar to bilateral filtering. only the comparison of each pixel with its immediate

neighbors is necessary.

Page 14: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1 Leonardo De-Maeztu, Arantxa Villanueva, Member, IEEE, and.

14

Method

A : Anisotropic diffusion

B : Geodesic diffusion

Page 15: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1 Leonardo De-Maeztu, Arantxa Villanueva, Member, IEEE, and.

15

Method.A(1/3)

Anisotropic diffusion

Page 16: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1 Leonardo De-Maeztu, Arantxa Villanueva, Member, IEEE, and.

16

Method.A(2/3)

Anisotropic diffusion

123

123 Euclidean distance between the values in the CIELab color space

Page 17: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1 Leonardo De-Maeztu, Arantxa Villanueva, Member, IEEE, and.

17

Method.A(3/3)

It is an iterative computer vision technique.[9]

Page 18: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1 Leonardo De-Maeztu, Arantxa Villanueva, Member, IEEE, and.

18

Method

A : Anisotropic diffusion

B : Geodesic diffusion

Page 19: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1 Leonardo De-Maeztu, Arantxa Villanueva, Member, IEEE, and.

19

Method.B(1/8)

Three principles Costs and weights are diffused so that the importance of

each cost value is known in each iteration. In each iteration, the costs and weights at each pixel are

accumulated. After the last iteration, all the support region information has been accumulated at each pixel.

To increase the efficiency of information diffusion and to avoid loops, turns in the direction of diffusion are penalized.

Page 20: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1 Leonardo De-Maeztu, Arantxa Villanueva, Member, IEEE, and.

20

Method.B(2/8)

Geodesic diffusion

Page 21: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1 Leonardo De-Maeztu, Arantxa Villanueva, Member, IEEE, and.

21

Method.B(3/8)

Each of the four positions inherits the costs and weights of each of the four direct neighbors of each pixel.

Page 22: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1 Leonardo De-Maeztu, Arantxa Villanueva, Member, IEEE, and.

22

Method.B(4/8)

Geodesic diffusion 123

123

123

Page 23: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1 Leonardo De-Maeztu, Arantxa Villanueva, Member, IEEE, and.

23

Method.B(5/8)

i = 0 right neighbors i = 1 lower neighbors i = 2 upper neighbors i = 3 left neighbors

Page 24: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1 Leonardo De-Maeztu, Arantxa Villanueva, Member, IEEE, and.

24

Method.B(6/8)

The cost and weight information derived from a direct neighbor is not returned to this neighbor.

Costs are only propagated with their full weights in the same direction of their propagation direction in the previous iteration.

Page 25: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1 Leonardo De-Maeztu, Arantxa Villanueva, Member, IEEE, and.

25

Method.B(7/8)

Geodesic diffusion 123

123

Page 26: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1 Leonardo De-Maeztu, Arantxa Villanueva, Member, IEEE, and.

26

Method.B(8/8)

At the end of the diffusion process, the DSI costs are normalized.

Thus, concluded, and the disparity map is then computed by selecting the lower cost disparity for each pixel WTA.

Page 27: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1 Leonardo De-Maeztu, Arantxa Villanueva, Member, IEEE, and.

27

Experimental Results(1/8)

Page 28: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1 Leonardo De-Maeztu, Arantxa Villanueva, Member, IEEE, and.

28

Experimental Results(2/8)

Page 29: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1 Leonardo De-Maeztu, Arantxa Villanueva, Member, IEEE, and.

29

Experimental Results(3/8)

Page 30: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1 Leonardo De-Maeztu, Arantxa Villanueva, Member, IEEE, and.

30

Experimental Results(4/8)

Page 31: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1 Leonardo De-Maeztu, Arantxa Villanueva, Member, IEEE, and.

31

Experimental Results(5/)

Page 32: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1 Leonardo De-Maeztu, Arantxa Villanueva, Member, IEEE, and.

32

Experimental Results(6/8)

Page 33: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1 Leonardo De-Maeztu, Arantxa Villanueva, Member, IEEE, and.

33

Experimental Results(7/8)

Page 34: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1 Leonardo De-Maeztu, Arantxa Villanueva, Member, IEEE, and.

34

Experimental Results(8/8)

Page 35: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1 Leonardo De-Maeztu, Arantxa Villanueva, Member, IEEE, and.

35

CUDA

CUDA implementation of our algorithm ran in less than 60 milliseconds for the Tsukuba stereo pair on a GeForce 480 GTX card.

Page 36: IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, VOL. 34, NO. 2, FEBRUARY 2012 1 Leonardo De-Maeztu, Arantxa Villanueva, Member, IEEE, and.

36

Q & A


Recommended