标题:A novel method for graph matching based on belief propagation
作者:Lin, Xue; Niu, Dongmei; Zhao, Xiuyang; Yang, Bo; Zhang, Caiming
作者机构:[Lin, Xue; Niu, Dongmei; Zhao, Xiuyang; Yang, Bo] Univ Jinan, Shandong Prov Key Lab Network Based Intelligent C, Jinan 250022, Shandong, Peoples R Chi 更多
通讯作者:Zhao, Xiuyang;Zhao, XY
通讯作者地址:[Zhao, XY]Univ Jinan, Shandong Prov Key Lab Network Based Intelligent C, Jinan 250022, Shandong, Peoples R China.
来源:NEUROCOMPUTING
出版年:2019
卷:325
页码:131-141
DOI:10.1016/j.neucom.2018.10.018
关键词:Graph matching; Energy minimization; Random sample consensus; Max-pooled; supports; Belief propagation; One-to-one match
摘要:Graph matching is a fundamental NP-problem in computer vision and pattern recognition. In this paper, we propose a robust approximate graph matching method. The match between two graphs is formulated as an optimization problem and a novel energy function that performs random sample consensus (RANSAC) checking on the max-pooled supports is proposed. Then a belief propagation(BP) algorithm, which can assemble the spatial supports of the local neighbors in the context of the given points, is used to minimize the energy function. To achieve the one-to-(at most)-one matching constraint, we present a method for removing bad matches based on the topological structure of the graphs. Experimental results demonstrate that the proposed method outperforms other state-of-the-art graph matching methods in matching accuracy. (C) 2018 Elsevier B.V. All rights reserved.
收录类别:EI;SCOPUS;SCIE
资源类型:期刊论文
原文链接:https://www.scopus.com/inward/record.uri?eid=2-s2.0-85055267566&doi=10.1016%2fj.neucom.2018.10.018&partnerID=40&md5=448ed015e1b9ed6c4eac94c0948508ec
TOP