Shape Similarity based on Combinatorial Maps and a Tree Pattern Kernel

Luc Brun (1) , Sébastien Bougleux (1) , François-Xavier Dupé (2) , Benoit Gaüzère (1) , Myriam Mokhtari (1)

1 Equipe Image - Laboratoire GREYC - UMR6072
2 LIF - Laboratoire d'informatique Fondamentale de Marseille - UMR 6166

While the skeleton of a 2D shape corresponds to a planar graph, its encoding by usual graph data struc- tures does not allow to capture its planar properties. Graph kernels may be defined on graph's encoding of the skeleton in order to define a similarity measure be- tween shapes. Such graph kernels are usually based on a decomposition of graphs into bags of walks or trails. These linear patterns do not allow to fully encode the structure of a skeleton on branching points, hence los- ing important informations about the shape. This paper aims to solve these two drawbacks by using an encoding of the skeleton taking explicitly into account the orien- tation of the plane and by decomposing the resulting graph model into both linear and nonlinear patterns.

Mots clés: shape analysis; combinatorial map; graph kernels.; graph kernels

Domaines: Vision par ordinateur et reconnaissance de formes [cs.CV]

Luc Brun, Sébastien Bougleux, François-Xavier Dupé, Benoit Gaüzère, Myriam Mokhtari. Shape Similarity based on Combinatorial Maps and a Tree Pattern Kernel. 21st International Conference on Pattern Recognition, Nov 2012, Tsukuba, Japan. pp.000-0000. ⟨hal-00768662⟩