Graph based recognition of isolated graphic symbols
Reeba Qureshi, Jean-Yves Ramel, Usman Ali, Hubert Cardot · Proceedings of the IEEE Symposium on Emerging Technologies, 2005. · 2005
Graphmatching methods areoften proposed tosolve thesymbols recognition problems. Finding anexact matchingbetweentwo graphsusinggraph isonmorphism techniques renmains uinresolved and knownto be NP-complete. In ourproposed technique, first, raster imageofeachgraphic symnbol isvectorized andquadrilaterals areconstructed by searching andselecting vectors having opposite direction andminimumdistance between their extremities. We utsequiadrilaterals and their neighbouring relations toconstruict alabelled graph. Todealwiththepresence ofnoise anddistortion, our approach isbased oninexact graph matching where eachvertex ofonegraphcanbeassociated tozero, oneormorevertices oftheother graph. Ituises a greedv incomplete search algorithn tofind thebest mapping between thevertices ofthetwvo graphs and compute a similarity score. Thestrength ofthis approach ismaxinmum struictural representation whichis invariant to distortion and affine transformations.