Spectral median graphs applied to graphical symbol recognition

Abstract

Generalized median graph is a general concept useful to capture the essential information of a set of graphs. In addition, spectral techniques can be used to obtain approximate solutions of graph matching problems in a reasonable time. In this work we use the novel concept of spectral median graph which takes advantage of both the median concept and the spectral techniques, to synthesize the representative of a set of graphical symbols. The results show that this concept represents appropriately the most important intra-class features, while rejecting small distortions and, for extension, it can be used to infer a prototype of a set of symbols. © Springer-Verlag Berlin Heidelberg 2006.