Consensus of Two Graph Correspondences Through a Generalisation of the Bipartite Graph Matching
One of the most important processes related to structural pattern recognition is to compare the involved objects through representing them as attributed graphs and using error-tolerant graph matching methods. To do so, it is needed a first step to extract the graphs given the original objects and deduct the whole attribute values of nodes and edges. Depending on the application, there are several methods to obtain these graphs and so, the object at hand can be represented by several graphs, not only with different nodes and edges but also with different attribute domains. In the case that we have several graphs to represent the same object, we can deduct several correspondences between graphs. In this work, we want to solve the problem of having these correspondences by exploding this diversity to announce a final correspondence in which the incongruences introduced in the graph extraction and also the graph matching could be reduced. We present a consensus method which, given two correspondences between two pairs of attributed graphs generated by separate entities and with different attribute domains, enounces a final correspondence consensus considering the existence of outliers. Our method is based on a generalisation of the Bipartite graph matching algorithm that minimises the Edit cost of the consensus correspondence while forcing (to the most) to be the mean correspondence of the two original correspondences.
KeywordsBipartite graph matching Graph correspondence Consensus correspondence
Unable to display preview. Download preview PDF.
- 2.Franek, F., Jiang, X., He, C.: Weighted Mean of a Pair of Clusterings. Pattern Analysis Applications (2012)Google Scholar
- 6.Serratosa, F.: Speeding up Fast Bipartite Graph Matching through a New Cost Matrix. International Journal of Pattern Recognition & Artificial Intelligence 29(2) (2015)Google Scholar
- 11.Papadimitriou, C., Steiglitz, K.: Combinatorial Optimization: Algorithms and Complexity. Dover Publications (1998)Google Scholar
- 14.Timofte, R., Van Gool, L.: Adaptive and Weighted Collaborative Representations for Image Classification. Pattern Recognition Letters (Available Online 2014)Google Scholar
- 17.Riesen, K., Bunke, H., Fischer, A.: Improving Graph Edit Distance Approximation by Centrality Measures. In: International Conference on Pattern Recognition (2014)Google Scholar
- 18.Cortés, X., Serratosa, F.: Learning Graph-Matching Edit-Costs based on the Optimality of the Oracle’s Node Correspondences. Pattern Recognition Letters (2015)Google Scholar
- 22.Solé, A., Serratosa, F., Sanfeliu, A.: On the Graph Edit Distance Cost: Properties and Applications. Intern. Journal of Pattern Recognition and Artificial Intell. 26(5) (2012)Google Scholar
- 25.Moreno-García, C.F., Serratosa, F., Cortés, X.: Iterative Versus Voting Method to Reach Consensus Given Multiple Correspondences of Two Sets. In: IbPRIA 2015 (2015)Google Scholar