DIGITAL CONTENT FEATURE AMOUNT EXTRACTION METHOD, DEVICE AND PROGRAM, AND COMPUTER READABLE RECORDING MEDIUM

摘要

PROBLEM TO BE SOLVED: To simultaneously improve identification accuracy and noise resistance in identifying identity/similarity of digital contents.

SOLUTION: The input contents are divided in a plurality of regions, and a region feature amount is calculated on each region. The region feature amount and region number of each region are stored as a set. With the region feature amount as a key, the sets of region feature amounts and region numbers of the stored regions are sorted, and the region numbers arranged in a sorted sequence are made sequence information. A plurality of regions are selected from the sequence information based on a specified sequence selection start position and selection distances, and the sets of the selected region numbers are made partial sequence information. A plurality of pieces of partial sequence information obtained by carrying out processing for acquiring the partial sequence information a plurality of times are output as the feature amount.

COPYRIGHT: (C)2007,JPO&INPIT