Progressively Decomposing Graph Matching
Existing approaches to graph matching mainly include two types, i.e., the Koopmans-Beckmann's QAP formulation (KB-QAP) and Lawler's QAP formulation (L-QAP). The former is advantageous in scalability but disadvantageous in generality, while the latter is exactly the opposite. In this paper,...
Main Authors: | , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
IEEE
2019-01-01
|
Series: | IEEE Access |
Subjects: | |
Online Access: | https://ieeexplore.ieee.org/document/8679956/ |
id |
doaj-9893c03b13704f0b85e4484c8b70032c |
---|---|
record_format |
Article |
spelling |
doaj-9893c03b13704f0b85e4484c8b70032c2021-03-29T22:17:26ZengIEEEIEEE Access2169-35362019-01-017453494535910.1109/ACCESS.2019.29089258679956Progressively Decomposing Graph MatchingJin-Gang Yu0https://orcid.org/0000-0003-2148-2726Lichao Xiao1Jiarong Ou2Zhifeng Liu3School of Automation Science and Engineering, South China University of Technology, Guangzhou, ChinaSchool of Automation Science and Engineering, South China University of Technology, Guangzhou, ChinaSchool of Automation Science and Engineering, South China University of Technology, Guangzhou, ChinaSchool of Automation Science and Engineering, South China University of Technology, Guangzhou, ChinaExisting approaches to graph matching mainly include two types, i.e., the Koopmans-Beckmann's QAP formulation (KB-QAP) and Lawler's QAP formulation (L-QAP). The former is advantageous in scalability but disadvantageous in generality, while the latter is exactly the opposite. In this paper, we present a novel graph matching method, called progressively decomposing graph matching (PDGM), which can simultaneously possess the merits of the scalability of KB-QAP and the generality of L-QAP. Our method is motivated by a key observation that, the matching accuracy of KB-QAP can be dramatically boosted by properly introducing a guidance term into the formulation. Based on this observation, the proposed PDGM method progressively incorporates edge affinity information into the optimization procedure of KB-QAP through a guidance term, which mainly involves two iterative steps, i.e., solving the guided KB-QAP and updating the guidance matrix. The extensive experiments on both synthetic data and real image datasets demonstrate that our method can outperform the state-of-the-art in terms of the robustness to noise/deformation and outliers, and the good balance between effectiveness and computational efficiency.https://ieeexplore.ieee.org/document/8679956/Graph matchingquadratic assignment problemprogressively decomposing graph matchingFrank-Wolfe algorithm |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Jin-Gang Yu Lichao Xiao Jiarong Ou Zhifeng Liu |
spellingShingle |
Jin-Gang Yu Lichao Xiao Jiarong Ou Zhifeng Liu Progressively Decomposing Graph Matching IEEE Access Graph matching quadratic assignment problem progressively decomposing graph matching Frank-Wolfe algorithm |
author_facet |
Jin-Gang Yu Lichao Xiao Jiarong Ou Zhifeng Liu |
author_sort |
Jin-Gang Yu |
title |
Progressively Decomposing Graph Matching |
title_short |
Progressively Decomposing Graph Matching |
title_full |
Progressively Decomposing Graph Matching |
title_fullStr |
Progressively Decomposing Graph Matching |
title_full_unstemmed |
Progressively Decomposing Graph Matching |
title_sort |
progressively decomposing graph matching |
publisher |
IEEE |
series |
IEEE Access |
issn |
2169-3536 |
publishDate |
2019-01-01 |
description |
Existing approaches to graph matching mainly include two types, i.e., the Koopmans-Beckmann's QAP formulation (KB-QAP) and Lawler's QAP formulation (L-QAP). The former is advantageous in scalability but disadvantageous in generality, while the latter is exactly the opposite. In this paper, we present a novel graph matching method, called progressively decomposing graph matching (PDGM), which can simultaneously possess the merits of the scalability of KB-QAP and the generality of L-QAP. Our method is motivated by a key observation that, the matching accuracy of KB-QAP can be dramatically boosted by properly introducing a guidance term into the formulation. Based on this observation, the proposed PDGM method progressively incorporates edge affinity information into the optimization procedure of KB-QAP through a guidance term, which mainly involves two iterative steps, i.e., solving the guided KB-QAP and updating the guidance matrix. The extensive experiments on both synthetic data and real image datasets demonstrate that our method can outperform the state-of-the-art in terms of the robustness to noise/deformation and outliers, and the good balance between effectiveness and computational efficiency. |
topic |
Graph matching quadratic assignment problem progressively decomposing graph matching Frank-Wolfe algorithm |
url |
https://ieeexplore.ieee.org/document/8679956/ |
work_keys_str_mv |
AT jingangyu progressivelydecomposinggraphmatching AT lichaoxiao progressivelydecomposinggraphmatching AT jiarongou progressivelydecomposinggraphmatching AT zhifengliu progressivelydecomposinggraphmatching |
_version_ |
1724191984788176896 |