23
Thinning & Distance Field Advisor Ku-Yaw Chang Speaker Jhen-Yu Yang

Thinning & Distance Field

  • Upload
    wray

  • View
    25

  • Download
    1

Embed Size (px)

DESCRIPTION

Thinning & Distance Field. Advisor : Ku-Yaw Chang Speaker : Jhen -Yu Yang. Outline. Introduction Method Method 1 Method 2 Reference. Introduction. Thinning To produce a skeleton Iteratively removing voxels from the boundary. Introduction. Distance field Find ridge points - PowerPoint PPT Presentation

Citation preview

Page 1: Thinning & Distance Field

Thinning & Distance Field

Advisor:Ku-Yaw Chang

Speaker: Jhen-Yu Yang

Page 2: Thinning & Distance Field

2

Outline Introduction Method

Method 1 Method 2

Reference

2011/12/29

Page 3: Thinning & Distance Field

3

Introduction Thinning

To produce a skeleton Iteratively removing voxels from the boundary

2011/12/29

Page 4: Thinning & Distance Field

4

Introduction Distance field

Find ridge points Connect them

Polyp

2011/12/29

Page 5: Thinning & Distance Field

5

Outline Introduction Method

Method 1 Method 2

Reference

2011/12/29

Page 6: Thinning & Distance Field

6

Method 1 Use two-subfield thinning algorithm

Extracting medial curves on 3D images Source : [C.-M. Ma et al., 2002]

2011/12/29

An original object and its skeleton.

Page 7: Thinning & Distance Field

7

In a 3D binary image Voxels are partitioned into two subsets

C (object voxel) Marked by

C’ (background voxel) Marked by

A voxel marked Don’t care Can match either C or C’

Method 1

2011/12/29

Page 8: Thinning & Distance Field

8

Method 1 Use the Templates (or masks)

To test each boundary voxel Remove the simple point

2011/12/29

Tested voxels

Page 9: Thinning & Distance Field

9

Let x be a C-voxel Simple point

Only one C-component in N*(x) x is adjacent to only one C-component

Method 1

2011/12/29

C-component = { a1, d1, b1, e1, δ(x) };

Page 10: Thinning & Distance Field

10

Voxel x is U-deletable With a C-neighbor β(x)

Method 1

2011/12/29

x can be deleted

( 1 )

( 2 )

( 3 )

C-component = { a, b, c, β(x) };

Page 11: Thinning & Distance Field

11

Method 1

2011/12/29

Voxels are partitioned into two subfields Two directly adjacent voxels

In different subfields Two diagonally adjacent voxels

In same subfield

Directly adjacent

Diagonally adjacent

Page 12: Thinning & Distance Field

12

Branches

Method 1

2011/12/29

A tree structure object and its skeleton.

A letter ‘A’ and its skeleton.

Page 13: Thinning & Distance Field

13

Outline Introduction Method

Method 1 Method 2

Reference

2011/12/29

Page 14: Thinning & Distance Field

14

Method 2 Using a distance field

Compute an object’s centerline Source : [I. Bitter et al., 2001]

2011/12/29

Colon and its skeleton. Dinosaur and its skeleton.

Page 15: Thinning & Distance Field

15

Method 2 Compute the distance

Each inside voxel to the boundary Recorded at each voxel

2011/12/29

(1)

Page 16: Thinning & Distance Field

16

Method 2

2011/12/29

Another DT case

Source Result

Page 17: Thinning & Distance Field

17

Compute gradient vector For each voxel position Requires reading of

neighboring voxels

Method 2

2011/12/29

Vector and its arrow

(2)

Page 18: Thinning & Distance Field

18

Six classes of regions Flag non-uniform gradient vectors

Directions are non-uniform

Method 2

2011/12/29

(3)GVF: Gradient Vector Field

Page 19: Thinning & Distance Field

19

Method 2 Connect flagged voxels

Pick a flagged voxel and flag the corresponding voxel Start and traverse Stop when another flagged voxel is reached

2011/12/29

(4)

Page 20: Thinning & Distance Field

20

Method 2

2011/12/29

Results

Lobster and its skeleton. Aorta and its skeleton.

Page 21: Thinning & Distance Field

21

Outline Introduction Method

Method 1 Method 2

Reference

2011/12/29

Page 22: Thinning & Distance Field

22

Reference Cherng-Min Ma, Shu-Yen Wan. A medial-surface oriented 3-d two-subfield

thinning algorithm. Pattern Recognition Letters 22 (2001) 1439-1446 Cherng-Min Ma, Shu-Yen Wan, Her-Kun Chang. Extracting medial curves on

3D images. Pattern Recognition Letters 23 (2002) 895-904 Ingmar Bitter, Arie E. Kaufman, Mie Sato.Penalized-Distance Volumetric

Skeleton Algorithm.IEEE Transaction On Visualization And Computer Graphics, Vol. 7, No. 3, July-September 2001

2011/12/29

Page 23: Thinning & Distance Field

23

The endThanks.

2011/12/29