Similar Subgraph Retrieving for Japanese Historical Document Search System
Hiroaki Nagaoka, Mitsuru Nakata, Qi‐Wei Ge, Makoto Yoshimura · ITC-CSCC :International Technical Conference on Circuits Systems, Computers and Communications · 2015
In this paper, we propose a retrieval method of similar subgraphs in order to obtain similar partial images used in Japanese historical documents image retrieval system. In our system, the structure of characters is represented by a feature graph, and the structure of a string of each line of Japanese historical documents is stored as a feature graph which is called “document graph”. The structure of a string to be searched is specified as a feature graph that is called “search graph”. A search graph contains an “essential graph”, which shows an indispensable structure included in search results. The processes to search similar partial images are as follows: (1) get subgraphs of document graphs that are isomorphic to the essential graph, and construct candidates of search result from these subgraphs; (2) remove improper candidates based on some conditions; (3) calculate the similarities between remaining candidates and the search graph (these remaining candidates are called “similar subgraph”); (4) clip the partial areas as the search results, which correspond to similar subgraphs with high similarity from Japanese historical document images.