shape recognition and retrieval based on edit distance and dynamic programming

34
Shape Recognition and Retrieval Based on Edit Distance and Dynamic Programming PAN Hongfei ( 潘潘潘 ), LIANG Dong ( 潘 潘 ), TANG Jun ( 潘 潘 ), WANG Nian ( 潘 潘 ), LI Wei ( 潘 潘 ) Presented by Kaidul Islam 0907016 1

Upload: kaidul-islam-sazal

Post on 30-Jun-2015

365 views

Category:

Technology


1 download

DESCRIPTION

Presentation on 2nd seminar, KUET

TRANSCRIPT

Page 1: Shape Recognition and Retrieval Based on Edit Distance and  Dynamic Programming

Shape Recognition and Retrieval Based on Edit Distance and

Dynamic Programming

PAN Hongfei ( 潘鸿飞 ), LIANG Dong ( 梁 栋 ), TANG Jun ( 唐 俊 ), WANG Nian ( 王 年 ), LI Wei ( 李 薇 )

Presented byKaidul Islam

0907016

1

Page 2: Shape Recognition and Retrieval Based on Edit Distance and  Dynamic Programming

Glimpse of this Paper

A shape recognition and retrieval algorithm How the algorithm approaches:- Extracting skeletal features(Medial axis

transform)- Transforming features into string of symbols- Similarity measurement(Edit Distance)- Shape identification(Dynamic programming) Analyzing public datasets for comparison with

other approaches.

2

Page 3: Shape Recognition and Retrieval Based on Edit Distance and  Dynamic Programming

Structure Shape Representation

- Shape skeleton - Skeleton feature representation

Computation of Similarity between Strings Shape matching Dynamic Programming Test and Analysis

- Matching Analysis - Comparisons

Future Work3

Page 4: Shape Recognition and Retrieval Based on Edit Distance and  Dynamic Programming

Shape Skeleton(1/2)

• Skeleton feature satisfy distance and curvature criteria– every point on the skeleton should be

equidistant from two different boundary point– distance between the skeleton point and the

boundary point should be less than the maximum distance between the center of gravity of the contour and the boundary points.

4

Page 5: Shape Recognition and Retrieval Based on Edit Distance and  Dynamic Programming

Shape Skeleton(2/2)

5

Page 6: Shape Recognition and Retrieval Based on Edit Distance and  Dynamic Programming

Structure Shape Representation

- Shape skeleton - Skeleton feature representation

Computation of Similarity between Strings Shape matching Dynamic Programming Test and Analysis

- Matching Analysis - Comparisons

Future Work6

Page 7: Shape Recognition and Retrieval Based on Edit Distance and  Dynamic Programming

Transformation from a skeleton to strings

7

Page 8: Shape Recognition and Retrieval Based on Edit Distance and  Dynamic Programming

Structure Shape Representation

- Shape skeleton - Skeleton feature representation

Computation of Similarity between Strings Shape matching Dynamic Programming Test and Analysis

- Matching Analysis - Comparisons

Future Work8

Page 9: Shape Recognition and Retrieval Based on Edit Distance and  Dynamic Programming

Edit Distance!

• Edit distance is a way of quantifying how dissimilar two strings are.

• Determine minimum number of operations to turn one string to another.

• Operation include –

- insertion

- Deletion

- Substitution9

Page 10: Shape Recognition and Retrieval Based on Edit Distance and  Dynamic Programming

Edit Distance(2/2)

10

Page 11: Shape Recognition and Retrieval Based on Edit Distance and  Dynamic Programming

Structure Shape Representation

- Shape skeleton - Skeleton feature representation

Computation of Similarity between Strings Shape matching Dynamic Programming Test and Analysis

- Matching Analysis - Comparisons

Future Work11

Page 12: Shape Recognition and Retrieval Based on Edit Distance and  Dynamic Programming

The tabulation , D(i, j) Seq2(j) A R T S

Seq1(i) 0 1 2 3 4

0

M 1

A 2

T 3

H 4

S 512

Page 13: Shape Recognition and Retrieval Based on Edit Distance and  Dynamic Programming

The tabulation , D(i, j) Seq2(j) A R T S

Seq1(i) 0 1 2 3 4

0 0

M 1

A 2

T 3

H 4

S 513

Page 14: Shape Recognition and Retrieval Based on Edit Distance and  Dynamic Programming

The tabulation , D(i, j) Seq2(j) A R T S

Seq1(i) 0 1 2 3 4

0 0 1

M 1

A 2

T 3

H 4

S 514

Page 15: Shape Recognition and Retrieval Based on Edit Distance and  Dynamic Programming

The tabulation , D(i, j) Seq2(j) A R T S

Seq1(i) 0 1 2 3 4

0 0 1 2

M 1

A 2

T 3

H 4

S 515

Page 16: Shape Recognition and Retrieval Based on Edit Distance and  Dynamic Programming

The tabulation , D(i, j) Seq2(j) A R T S

Seq1(i) 0 1 2 3 4

0 0 1 2 3 4

M 1 1

A 2 2

T 3 3

H 4 4

S 5 516

Page 17: Shape Recognition and Retrieval Based on Edit Distance and  Dynamic Programming

The tabulation , D(i, j) Seq2(j) A R T S

Seq1(i) 0 1 2 3 4

0 0 1 2 3 4

M 1 1 1

A 2 2

T 3 3

H 4 4

S 5 517

Page 18: Shape Recognition and Retrieval Based on Edit Distance and  Dynamic Programming

The tabulation , D(i, j) Seq2(j) A R T S

Seq1(i) 0 1 2 3 4

0 0 1 2 3 4

M 1 1 1 2

A 2 2

T 3 3

H 4 4

S 5 518

Page 19: Shape Recognition and Retrieval Based on Edit Distance and  Dynamic Programming

The tabulation , D(i, j) Seq2(j) A R T S

Seq1(i) 0 1 2 3 4

0 0 1 2 3 4

M 1 1 1 2 3 4

A 2 2 1 2 3 4

T 3 3

H 4 4

S 5 519

Page 20: Shape Recognition and Retrieval Based on Edit Distance and  Dynamic Programming

The tabulation , D(i, j) Seq2(j) A R T S

Seq1(i) 0 1 2 3 4

0 0 1 2 3 4

M 1 1 1 2 3 4

A 2 2 1 2 3 4

T 3 3 2 2 2 3

H 4 4

S 5 520

Page 21: Shape Recognition and Retrieval Based on Edit Distance and  Dynamic Programming

The tabulation , D(i, j) Seq2(j) A R T S

Seq1(i) 0 1 2 3 4

0 0 1 2 3 4

M 1 1 1 2 3 4

A 2 2 1 2 3 4

T 3 3 2 2 2 3

H 4 4 3 3 3 3

S 5 5 4 4 4 321

Page 22: Shape Recognition and Retrieval Based on Edit Distance and  Dynamic Programming

The traceback Seq2(j) A R T S

Seq1(i) 0 1 2 3 4

0 0 1 2 3 4

M 1 1 1 2 3 4

A 2 2 1 2 3 4

T 3 3 2 2 2 3

H 4 4 3 3 3 3

S 5 5 4 4 4 322

Page 23: Shape Recognition and Retrieval Based on Edit Distance and  Dynamic Programming

Similarity Cost by Dynamic Programming

23

Page 24: Shape Recognition and Retrieval Based on Edit Distance and  Dynamic Programming

Structure Shape Representation

- Shape skeleton - Skeleton feature representation

Computation of Similarity between Strings Shape matching Dynamic Programming Test and Analysis

- Matching Analysis - Comparisons

Future Work24

Page 25: Shape Recognition and Retrieval Based on Edit Distance and  Dynamic Programming

Matching Analysis(1/2)Shape sets from the MPEG-7 Shape Silhouette

Database

25

Page 26: Shape Recognition and Retrieval Based on Edit Distance and  Dynamic Programming

Matching Analysis(2/2)Comparison of results for various values of k

26

Page 27: Shape Recognition and Retrieval Based on Edit Distance and  Dynamic Programming

Structure Shape Representation

- Shape skeleton - Skeleton feature representation

Computation of Similarity between Strings Shape matching Dynamic Programming Test and Analysis

- Matching Analysis - Comparisons

Future Work27

Page 28: Shape Recognition and Retrieval Based on Edit Distance and  Dynamic Programming

Comparisons(1/3)

Sample shape sets from the MPEG-7 Shape Silhouette Database(1)

Sample shape sets from Plant Species Database(2)

28

Page 29: Shape Recognition and Retrieval Based on Edit Distance and  Dynamic Programming

Comparisons(2/3)

Retrieval rates for Silhouette Database(1)

Retrieval rates for Silhouette Database(2)

29

Page 30: Shape Recognition and Retrieval Based on Edit Distance and  Dynamic Programming

Comparisons(3/3)

30

Page 31: Shape Recognition and Retrieval Based on Edit Distance and  Dynamic Programming

Comparisons(3/3)

31

Page 32: Shape Recognition and Retrieval Based on Edit Distance and  Dynamic Programming

Structure Shape Representation

- Shape skeleton - Skeleton feature representation

Computation of Similarity between Strings Shape matching Dynamic Programming Test and Analysis

- Matching Analysis - Comparisons

Future Work32

Page 33: Shape Recognition and Retrieval Based on Edit Distance and  Dynamic Programming

Future Work

• Improve the algorithm

• Introducing other features –

- Texture

- Movement

33

Page 34: Shape Recognition and Retrieval Based on Edit Distance and  Dynamic Programming

34