Image retrieval with graph kernel on regions.

Justine Lebrun (1) , Sylvie Philipp-Foliguet (1) , Philippe-Henri Gosselin (2)

1 ETIS - UMR 8051 - Equipes Traitement de l'Information et Systèmes
2 MIDI - Multimedia Indexation and Data Integration

In the framework of the interactive search in image databases, we are interested in similarity measures able to learn during the search and usable in real-time. Im- ages are represented by adjacency graphs of fuzzy re- gions. In order to compare attributed graphs, we em- ploy kernels on graphs built on sets of paths. In this pa- per, we introduce a fast kernel function whose similar- ity is based on several matches. We also introduce new features for edges in the graph. Experiments on a spe- cific database having objects with heterogeneous back- grounds show the performance of our object retrieval technique.

Domaines: Machine Learning [stat.ML]

Justine Lebrun, Sylvie Philipp-Foliguet, Philippe-Henri Gosselin. Image retrieval with graph kernel on regions.. International Conference on Pattern Recognition, Dec 2008, United States. pp.1. ⟨hal-00520300⟩