Publications

Detailed Information

Graph and Hypergraph Matching in Computer Vision: Algorithms and Analysis : 컴퓨터비전을 위한 그래프정합과 고차그래프정합: 새로운 알고리즘과 분석에 관한 연구

DC Field Value Language
dc.contributor.advisor이경무-
dc.contributor.author이정민-
dc.date.accessioned2017-07-13T07:17:02Z-
dc.date.available2017-07-13T07:17:02Z-
dc.date.issued2016-08-
dc.identifier.other000000137035-
dc.identifier.urihttps://hdl.handle.net/10371/119217-
dc.description학위논문 (박사)-- 서울대학교 대학원 : 전기·컴퓨터공학부, 2016. 8. 이경무.-
dc.description.abstractEstablishing image feature correspondences is fundamental problem in computer vision and machine learning research fields. Myriad of graph matching algorithms have been proposed to tackle this problem by regarding correspondence problem as a graph matching problem. However, the graph matching problem is challenging
since there are various types of noises in real world scenario
-
dc.description.abstracte.g., non-rigid motion, view-point change, and background clutter. The objective of this dissertation is
to propose robust graph matching algorithms for feature correspondence task in computer vision and to investigate an effective graph matching strategy.
For the purpose, at first, two robust simulation based graph matching algorithms are introduced: the one is based on Random Walks simulation and the other is based
on Markov Chain Monte Carlo sampling simulation. Secondly, two different graph matching formulations and their transformal relation are studied since equivalence
between two formulations are not well studied in graph matching fields. It is demonstrated that conventional graph matching algorithms can solve both types of formulations by proposing conversion principle between two formulations. Finally, these whole statements are extended into hypergraph matching problem by introducing
two robust hypergraph matching algorithms which are based on Random Walks and Markov Chain Monte Carlo, by relating two different hypergraph matching formulations, and by reinterpreting previous hypergraph matching algorithms into their counterpart formulations. Throughout chapters in this dissertation, comparative and extensive experiments verify characteristics of formulations, transformal relations, and algorithms. Synthetic graph matching problems as well as real image feature
correspondence problems are performed in various and severe noise conditions.
-
dc.description.tableofcontentsChapter 1 Introduction 1
1.1 Graph Matching Problem 1
1.1.1 Graph Matching for Computer Vision 1
1.1.2 Graph Matching Formulation 2
1.1.3 Extension to Hypergraph Matching 5
1.2 Outline of Dissertation 6

Chapter 2 Graph Matching via Random Walks 9
2.1 Introduction 9
2.1.1 Related Works 10
2.2 Problem Formulation 12
2.2.1 Graph Matching Formulation 12
2.2.2 Hypergraphs Matching Formulation 13
2.3 Graph Matching via Random Walks 16
2.3.1 Random Walks for Graph Matching 16
2.3.2 Reweighting Jumps for Graph Matching 19
2.4 Hypergraph Matching via Random Walks 22
2.4.1 Hypergraph Random Walks 22
2.4.2 Reweighting Jumps for Hypergraph Matching 23
2.5 Experiments 26
2.5.1 Random Graph Matching 27
2.5.2 Synthetic Point Matching 34
2.5.3 Image Sequence Matching 37
2.5.4 Image Feature Matching 39
2.6 Conclusion 44

Chapter 3 Graph Matching via Markov Chain Monte Carlo 45
3.1 Introduction 45
3.2 Graph Matching Formulation 47
3.3 Algorithm 49
3.3.1 State Transition 49
3.3.2 Energy Formulation 49
3.3.3 Data-Driven Proposal 51
3.4 Hypergraph Extension 53
3.4.1 Hypergraph Matching Problem 53
3.4.2 Energy Formulation & Data-Driven Proposal 54
3.5 Experiment 54
3.5.1 Random Graph Matching Problem 54
3.5.2 Random Hypergraph Matching Problem 58
3.6 Conclusion 59

Chapter 4 Graph and Hypergraph Matching Revisited 63
4.1 Introduction 63
4.2 Related Works 65
4.3 Two Types of Formulations 66
4.3.1 Adjacency-based Formulation 67
4.3.2 Affinity-based Formulation 69
4.3.3 Relation between Two Formulations 70
4.4 Affinity Measures 72
4.5 Existing Methods & Re-interpretations 74
4.5.1 Spectral Matching 74
4.5.2 Integer Projected Fixed Point 75
4.5.3 Reweighted Random Walks Matching 76
4.5.4 Factorized Graph Matching 77
4.6 High-order Methods & Reinterpretations 78
4.6.1 Hypergraph Matching by Zass and Shashua 81
4.6.2 SVD-based Hypergraph Matching 82
4.6.3 Tensor Power Iteration based Hypergraph Matching 82
4.6.4 Reweighted Random Walks for Hypergraph Matching 83
4.6.5 Discrete Hypergraph Matching 85
4.7 Experiments & Comparison 85
4.8 Conclusion 102

Chapter 5 Conclusion 105
5.1 Summary and Contribution of Dissertation 105
5.2 Future Works 107

Bibliography 109

국문 초록 117
-
dc.formatapplication/pdf-
dc.format.extent12345758 bytes-
dc.format.mediumapplication/pdf-
dc.language.isoen-
dc.publisher서울대학교 대학원-
dc.subjectGraph Matching-
dc.subjectHypergraph Matching-
dc.subjectGraph Matching Formulations-
dc.subjectMarkov chain Monte Carlo-
dc.subjectData-Driven-
dc.subjectRandom Walks-
dc.subject.ddc621-
dc.titleGraph and Hypergraph Matching in Computer Vision: Algorithms and Analysis-
dc.title.alternative컴퓨터비전을 위한 그래프정합과 고차그래프정합: 새로운 알고리즘과 분석에 관한 연구-
dc.typeThesis-
dc.description.degreeDoctor-
dc.citation.pages119-
dc.contributor.affiliation공과대학 전기·컴퓨터공학부-
dc.date.awarded2016-08-
Appears in Collections:
Files in This Item:

Altmetrics

Item View & Download Count

  • mendeley

Items in S-Space are protected by copyright, with all rights reserved, unless otherwise indicated.

Share