The Community for Technology Leaders
Pattern Recognition, International Conference on (2002)
Quebec City, QC, Canada
Aug. 11, 2002 to Aug. 15, 2002
ISSN: 1051-4651
ISBN: 0-7695-1695-X
pp: 40176
Antonio Robles-Kelly , University of York
Edwin. R. Hancock , University of York
ABSTRACT
This paper describes a spectral method for graph-matching. We adopt a graphical models viewpoint in which the graph adjacency matrix is taken to represent the transition probability matrix of a Markov chain. The node-order of the steady state random walk associated with this Markov chain is determined by the co-efficent order of the leading eigenvector of the adjacency matrix. We match nodes in different graphs by aligning their sequence order in the steady-state walk. The method proceeds from the nodes with the largest leading eigenvector co-efficient. We develop a brushfir e search method to assign correspondences between nodes using the rank-order of the eigenvector coefficients in first-order neighbourhoods of the graphs. We demonstrate the utility of the new graph-matching method on both synthetic and real graphs.
INDEX TERMS
null
CITATION

A. Robles-Kelly and E. R. Hancock, "A Graph-Spectral Approach to Correspondence Matching," Pattern Recognition, International Conference on(ICPR), Quebec City, QC, Canada, 2002, pp. 40176.
doi:10.1109/ICPR.2002.1047426
95 ms
(Ver 3.3 (11022016))