This paper describes the fast matching of hierarchical attributed relational graphs with the aim of applying the method to similarity-based image retrieval. Best-first search A* algorithm, admissible heuristic function, and maximum permissible cost are proposed to speed up the computation of graph matching. By means of these methods, the average computation time of graph matching speeds up about three times faster than that of using conventional methods, which speed-up reduces the retrieval time into only five seconds in the case of retrieving top 10 similar graphs from the database of 1027 graphs.

      @InProceedings{ kt:miru96,
      author = {Asanobu KITAMOTO and Mikio TAKAGI},
      title = {Fast Matching of Hierarchical Attributed Relational Graphs for an Application to Similarity-Based Image Retrieval},
      booktitle = {Meeting on Image Recognition and Understanding (MIRU'96)},
      volume = {II},
      pages = {331-336},
      year = 1996,
      month = 7,
      doi = {},
      note = { (in Japanese)},
      }