+ All Categories
Home > Documents > The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine...

The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine...

Date post: 12-Jan-2016
Category:
Upload: shanon-mavis-allen
View: 221 times
Download: 1 times
Share this document with a friend
Popular Tags:
39
The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague [email protected] Courtesy Martin Urban (the talk is based on his presentation) and Jana Kostkova, Jan Kybic, Jiri Matas, Radim Sara
Transcript
Page 1: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

The Correspondence Problemand “Interest Point” Detection

Václav HlaváčCenter for Machine Perception

Czech Technical University [email protected]

Courtesy Martin Urban (the talk is based on his presentation) and Jana Kostkova, Jan Kybic, Jiri Matas, Radim Sara

Page 2: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

2

1. The Correspondence Problem (CP) lies at the core of a number of computer vision problems:

• tracking (correspondence in consecutive frames• narrow-baseline stereo• wide-baseline stereo• egomotion (camera motion) estimation • motion segmentation in sequences with a moving camera• recognition, categorisation, …

2. Demo of typical applications3. The CP is commonly solved by robust matching of

“Interest Points”4. “Interest Points” are regions with distinguishing

property, allowing there detection in a viewpoint and illumination invariant manner

5. Harris “Corner” (HC) Detection Algorithm• HC are rotation and translation invariant interest points• HC are interest points most commonly used in tracking and stereo

Lecture Overview

Page 3: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

3

“Ideal” solution: a pixel to pixel mapping from A to B (B ⋃ NULL)

What it is “The correspondence problem”?

A B

Ex. result: x-shift

Page 4: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

Applications: 3D Reconstruction

Page 5: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

3D Reconstruction

Page 6: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

Camera motion tracking ⇒ image stabilization

original stabilized

original stabilized

Page 7: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

Camera motion tracking ⇒ 3D animation

Page 8: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

Motion keying / segmentation

input sequence

Background sequence Foreground sequence

Page 9: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

Motion Detection

Input:

Output:

Page 10: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

Medical imaging – image registration

from the atlas

test slice

deform. field

before registration

after

Page 11: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

11

“Ideal” solution: a pixel to pixel mapping from A to B

The correspondence problem

A B

How to do it?

Page 12: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

12

The correspondence problem

A BIntuitive approach:

For ∀ pxl ∈ A, find a pxl ∈ B with the most similar neighbourhood.

Problems: - How to measure similarity of image patches? - undistinguishable regions (e.g. texture-less) - not surjective map (onto) due to occlusions - not bijective map (one to one) due to scale changes

- huge data ⇒ very hard / impossible to recover the mapping by direct

minimisation

Page 13: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

13

The correspondence problem

Conclusion:It is very hard / impossible to recover dense pxl 2 pxl mapping between two images.

Solution: 1. Recover the correspondence relation just between several well distinguished image features (interest points / corners / Harris points). 2. Estimate multi-view transformation (e.g. epipolar geometry, camera motion).

3. Having epipolar geometry, try to find dense correspondences.

Page 14: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

14

Corner Detection: Introduction

Corner detector detects points with distinguished neighbourhood(*) well suited for matching verification.

undistinguished patches:

distinguished patches:

Page 15: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

15

-

Demo of a point + with well distinguished neighbourhood.

+

> 0

Corner Detection: Introduction

Page 16: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

16

-

Demo of a point + with well distinguished neighbourhood.

+

> 0

Corner Detection: Introduction

Page 17: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

17

-

Demo of a point + with well distinguished neighbourhood.

+

> 0

Corner Detection: Introduction

Page 18: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

18

-

Demo of a point + with well distinguished neighbourhood.

+

> 0

Corner Detection: Introduction

Page 19: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

19

-

Demo of a point + with well distinguished neighbourhood.

+

> 0

Corner Detection: Introduction

Page 20: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

20

-

Demo of a point + with well distinguished neighbourhood.

+

> 0

Corner Detection: Introduction

Page 21: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

21

-

Demo of a point + with well distinguished neighbourhood.

+

> 0

Corner Detection: Introduction

Page 22: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

22

-

Demo of a point + with well distinguished neighbourhood.

+

> 0

Corner Detection: Introduction

Page 23: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

23

Example of detected points

Corner Detection: Introduction

Page 24: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

24

Corner Detection: Basic principle

undistinguished patches: distinguished patches:

Image gradients ∇I(x,y) of undist. patches are (0,0) or have only one principle component.

Image gradients ∇I(x,y) of dist. patches have two principle components.

⇒ rank ( ∑ ∇I(x,y)* ∇I(x,y) ⊤ ) = 2

Page 25: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

25

Algorithm (C. Harris, 1988)

1. filter the image by gaussian (2x 1D convolution), sigma_d

2. compute the intensity gradients ∇I(x,y), (2x 1D conv.)

3. for each pixel and given neighbourhood, sigma_i:

- compute auto-correlation matrix

A = ∑ ∇I(x,y)* ∇I(x,y) ⊤

- and evaluate the response function R(A):

R(A) >> 0 for rank(A)=2, R(A) → 0 for rank(A)<2

4. choose the best candidates (non-max suppression and thresholding)

Page 26: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

26

Corner Detection: Algorithm (R. Harris, 1988)

Harris response function R(A):

R(A) = det (A) – k*trace 2(A) ,

[lamda1,lambda2] = eig(A)

Page 27: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

27

Corner Detection: Algorithm (R. Harris, 1988)

Algorithm properties:

+ “invariant” to 2D image shift and rotation

+ invariant to shift in illumination

+ “invariant” to small view point changes

+ low numerical complexity

- not invariant to larger scale changes

- not invariant to high contrast changes

- not invariant to bigger view point changes

Page 28: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

28

Corner Detection: Algorithm (C. Harris, 1988)

Exp.: Harris points and view point change

Page 29: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

29

Corner Detection: Harris points versus sigma_d and sigma_i

Sigma_I →

↑Sigma_d

Page 30: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

30

Corner Detection: Application

Algorithm:

1. Corner detection

2. Tentative correspondences- by comparing similarity of the corner neighb. in the searching window (e.g. cross-correlation)

3. Camera motion geometry estimation (e.g. by RANSAC)- finds the motion geometry and consistent correspondences

4. 3D reconstruction- triangulation, bundle adjustment

3D camera motion tracking / 3D reconstruction

Page 31: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

31

Corner Detection: Camera Tracking Application - Boujou

Input sequence

Page 32: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

32

Corner Detection: Camera Tracking Application - Boujou

Harris points

Page 33: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

33

Corner Detection: Camera Tracking Application

points consistent with 3D camera motion

Page 34: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

34

Corner Detection: Camera Tracking Application

3D points and 3D camera motion

Page 35: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

35

Corner Detection: Camera Tracking Application

3D animation

Page 36: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

36

Corner Detection: Camera Tracking Application - Boujou

Input sequence

Page 37: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

37

Corner Detection: Camera Tracking Application - Boujou

Harris points

Page 38: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

38

Corner Detection: Camera Tracking Application - Boujou

points consistent with 3D camera motion

Page 39: The Correspondence Problem and “Interest Point” Detection Václav Hlaváč Center for Machine Perception Czech Technical University Prague hlavac@fel.cvut.cz.

39

for the presentation I copied almost done presntation• Martin Urban

for used demo images and software:• Jana Kostková, CMP - (slide 3: images, disparity map)• Radim Šára, CMP - (slide 5: images, 3D face reconstruction

demo)• Jan Kybic, CMP - (slide 10: medical image registration)

• 2d3 – (slide 31-38: Boujou Demo, slide 6: img. sequences) • ImagineerSystems – (slide 8: MoKey Demo, slide 8: img.

sequences)• The Pixel Farm – (slide 31-33: img. sequence)

Acknowledgements


Recommended