Japanese / English


論文の言語 英語
著者 Koichi KISE, Satoshi YOKOTA, and Akira SHIOZAKI
論文名 Copyright Protection of Images Based on Large-Scale Image Recognition
論文誌名 IEICE Technical Report
Vol. 107
No. 281
発表番号 PRMU2007-91
ページ pp.3-8
年月 2007年10月
要約 This paper presents a method of copyright protection for digital still images based on their recognition with a large-scale image database. For the purpose of copyright protection of images, digital watermarking has been extensively studied for more than a decade. However,it is still difficult to achieve enough robustness against alternations such as StirMark attacks. In order to solve this problem, we propose a method of copyright protection based on new image recognition technologies that are tolerant to various image distortions including geometric transformations. The image recognition enables us to find a corresponding image from at least tens of thousands of reference images to be protected in response to a query image that undergoes StirMark attacks and digital watermarking. From experimental results using 10,000 images in the database, we have confirmed that the method is capable of recognizing 99.5\% of query images only within 216 ms / query (excluding feature extraction).