A String Matching Approach for Visual Retrieval and Classification Mei-Chen Yeh* and Kwang-Ting...

22
A String Matching Approach for Visual Retrieval and Classification Mei-Chen Yeh* and Kwang-Ting Cheng Learning-Based Multimedia Lab Department of Electrical and Computer Engineering University of California, Santa Barbara
  • date post

    22-Dec-2015
  • Category

    Documents

  • view

    225
  • download

    1

Transcript of A String Matching Approach for Visual Retrieval and Classification Mei-Chen Yeh* and Kwang-Ting...

Page 1: A String Matching Approach for Visual Retrieval and Classification Mei-Chen Yeh* and Kwang-Ting Cheng Learning-Based Multimedia Lab Department of Electrical.

A String Matching Approach for Visual Retrieval and ClassificationMei-Chen Yeh* and Kwang-Ting Cheng

Learning-Based Multimedia LabDepartment of Electrical and Computer EngineeringUniversity of California, Santa Barbara

Page 2: A String Matching Approach for Visual Retrieval and Classification Mei-Chen Yeh* and Kwang-Ting Cheng Learning-Based Multimedia Lab Department of Electrical.
Page 3: A String Matching Approach for Visual Retrieval and Classification Mei-Chen Yeh* and Kwang-Ting Cheng Learning-Based Multimedia Lab Department of Electrical.

Representation Based on Local Features

Belongie et al.; Berg. et al.; Ling & Jacobs Ozkan & Duyulu; Bicego et al.; Zou et al.

Page 4: A String Matching Approach for Visual Retrieval and Classification Mei-Chen Yeh* and Kwang-Ting Cheng Learning-Based Multimedia Lab Department of Electrical.

Matching

• Features are unordered• Similarity is measured only on a feature subset

Page 5: A String Matching Approach for Visual Retrieval and Classification Mei-Chen Yeh* and Kwang-Ting Cheng Learning-Based Multimedia Lab Department of Electrical.

Alternative Representation?

],...,,[ 21 mxxx

Globally ordered and locally unordered!

• Order carries useful information• Incorporate order into representation

Page 6: A String Matching Approach for Visual Retrieval and Classification Mei-Chen Yeh* and Kwang-Ting Cheng Learning-Based Multimedia Lab Department of Electrical.

Matching

Order of features is constrained Sequences may not be of equal lengths Tolerate errors Incorporate the ground distance Measure similarity based on both matched and

unmatched features

],...,,[ 21 mxxx

],...,,[ 21 nyyy

Page 7: A String Matching Approach for Visual Retrieval and Classification Mei-Chen Yeh* and Kwang-Ting Cheng Learning-Based Multimedia Lab Department of Electrical.

Outline

Introduction Representation based on ordered features Matching criteria

Approach Experiments

Shape retrieval Scene recognition

Conclusion and future work

Page 8: A String Matching Approach for Visual Retrieval and Classification Mei-Chen Yeh* and Kwang-Ting Cheng Learning-Based Multimedia Lab Department of Electrical.

Approximate String Matching

],...,,[ 21 mxxx

i

icYXd min),(

],...,,[ 21 nyyy

The distance is defined as the minimal cost of operations that transform X into Y

Page 9: A String Matching Approach for Visual Retrieval and Classification Mei-Chen Yeh* and Kwang-Ting Cheng Learning-Based Multimedia Lab Department of Electrical.

Edit Distance

Three operations Insertion δ (ε, a) Deletion δ (a, ε) Substitution δ (a, b)

Application dependentIntegrates the ground distance

Example: Matching two shapes• Shape Context Descriptor • χ2 distance d(∙,∙)

δ (ε, a) = d ( 0, a ) δ (a, ε) = d ( a, 0 )δ (a, b) = d ( a, b)

2

Page 10: A String Matching Approach for Visual Retrieval and Classification Mei-Chen Yeh* and Kwang-Ting Cheng Learning-Based Multimedia Lab Department of Electrical.

Edit Distance (cont.)

Metric if each operation cost is also a metric

Preserve the ordering The edit distance reflects

The similarity of the corresponding features by substitution

The dissimilarity of unmatched features by insertion/deletion

Easy to incorporate the ground distance No need to create a visual alphabet

Page 11: A String Matching Approach for Visual Retrieval and Classification Mei-Chen Yeh* and Kwang-Ting Cheng Learning-Based Multimedia Lab Department of Electrical.

Computing the edit distance

)}.,()1,(

),,(),1(

),,()1,1(min{),(

j

i

ji

yjiD

xjiD

yxjiDjiD

Computation O(mn) Space O(min(m, n))

Page 12: A String Matching Approach for Visual Retrieval and Classification Mei-Chen Yeh* and Kwang-Ting Cheng Learning-Based Multimedia Lab Department of Electrical.

Alignment of Two Sequences

Cyclic sequences

Search for a pattern X in a duplicate string Y-Yy1 y1yn yn… …

0 0 0 0 0 0 0 0 0 0 0 0 0

x1

xm min { Dm, j }

Page 13: A String Matching Approach for Visual Retrieval and Classification Mei-Chen Yeh* and Kwang-Ting Cheng Learning-Based Multimedia Lab Department of Electrical.

Outline

Introduction Representation based on ordered features Matching criteria

Approach Experiments

Shape retrieval Scene recognition

Conclusion and future work

Page 14: A String Matching Approach for Visual Retrieval and Classification Mei-Chen Yeh* and Kwang-Ting Cheng Learning-Based Multimedia Lab Department of Electrical.

Shape Retrieval

MPEG-7 Core Experiment CE-Shape-1 part B 1400 shapes, 70 categories Bull’s eye test# {correct hits in top 40} # {total possible hits}

Representation Uniformly sample 100 points 60-d (5 bins, 12 orientations) shape context

descriptor χ2 distance, δ (ε, a) = δ (a, ε) = 0.5

Page 15: A String Matching Approach for Visual Retrieval and Classification Mei-Chen Yeh* and Kwang-Ting Cheng Learning-Based Multimedia Lab Department of Electrical.

Shape Retrieval

71.81%

76.51%

82.46%84.29%

0.65

0.7

0.75

0.8

0.85

0.9

1 2 3 4

Same features (shape context descriptor)Different matching methods

Bipartite SC+TPS[Belongie et al.]

COPAP [Scott & Nowak]

Ours

73.51%

84.29% 85.40%

0.65

0.7

0.75

0.8

0.85

0.9

1 2 3

Different features

Fourier Descriptors

Ours(SC+ASM)

IDSC+DP[Ling & Jacobs]

• Find the alignment during matching• No need for any transformation model

Page 16: A String Matching Approach for Visual Retrieval and Classification Mei-Chen Yeh* and Kwang-Ting Cheng Learning-Based Multimedia Lab Department of Electrical.

Scene Recognition

Scene dataset [Lazebnik et al.] 15 categories, 4485 images, 200-400 images per

category 100 images for training, rest for testing

Representation Spatial pyramid representation Harris-Hessian-Laplace detector + SIFT features 200 visual words

Classification SVM with specified kernel values

Page 17: A String Matching Approach for Visual Retrieval and Classification Mei-Chen Yeh* and Kwang-Ting Cheng Learning-Based Multimedia Lab Department of Electrical.

Spatial Pyramid Representation

• Level-2 partitioning (16 bags-of-features) achieves the best performance• Each image is represented by 16 bags of features

Page 18: A String Matching Approach for Visual Retrieval and Classification Mei-Chen Yeh* and Kwang-Ting Cheng Learning-Based Multimedia Lab Department of Electrical.

Scene Recognition

Spatial Pyramid Matching : bag-to-bag matchingOur matching method could allow matching across bags

63.30%

69.21%

66.42%

0.6

0.62

0.64

0.66

0.68

0.7

1 2 3

Using the L-2 partitioning alone and the same ground distance χ2

SPM[Lazebnik et al.]

Bipartite Ours

Page 19: A String Matching Approach for Visual Retrieval and Classification Mei-Chen Yeh* and Kwang-Ting Cheng Learning-Based Multimedia Lab Department of Electrical.

back

86.74% 76.73% 95.48%

48.71% 78.69% 55.87%

35.21% 67.74% 52.65%

58.52% 73.91% 65.63%

46.55% 53.76% 77.72%

Page 20: A String Matching Approach for Visual Retrieval and Classification Mei-Chen Yeh* and Kwang-Ting Cheng Learning-Based Multimedia Lab Department of Electrical.

Conclusion

A globally ordered and locally unordered representation for visual data

Approximate String Matching for measuring the similarity between such representations Order is considered Naturally integrates the ground distance between

features Similarity is derived based on both matched and

unmatched features

Page 21: A String Matching Approach for Visual Retrieval and Classification Mei-Chen Yeh* and Kwang-Ting Cheng Learning-Based Multimedia Lab Department of Electrical.

Future Work

Image registration by using correspondences found by implementing this approach.

Video retrieval and video copy detection by exploring the local alignment ability of string matching approach.

Page 22: A String Matching Approach for Visual Retrieval and Classification Mei-Chen Yeh* and Kwang-Ting Cheng Learning-Based Multimedia Lab Department of Electrical.

Thank you

More information:http://lbmedia.ece.ucsb.edu