VISUAL KEYWORD SEARCH DEVICE AND VISUAL KEYWORD SEARCH METHOD

摘要

PROBLEM TO BE SOLVED: To provide a technique for reducing a search time in visual keyword search.;SOLUTION: When query data for searching a neighbor vocabulary is received, based on the query data, a leaf node is searched using a tree structure index. Then distance q between the query data and a central vocabulary corresponding to the searched leaf node, is calculated. In addition, difference data Si between distance di between the central vocabulary and plural neighbor vocabularies and the distance q, is calculated. A sort list, in which the plural neighbor vocabularies are sorted in order from one with the lowest absolute value of the calculated difference data Si to one with the highest absolute value, is created. In the order of the sort list, predetermined recursive processing followed by distance calculation is performed, sequentially, and nearest neighbor vocabulary candidates are updated. In a middle way of the sort list, the recursive processing is finished for specifying a proper nearest neighbor vocabulary. Therefore, distance calculation of the following neighbor vocabularies is omitted.;COPYRIGHT: (C)2016,JPO&INPIT