IRIW: Image Retrieval based Image Watermarking for Large-Scale Image Databases

by Jong Yun Jun, Kunho Kim, Jae-Pil Heo, and Sung-eui Yoon,

(To appear) 10th International Workshop on Digital-forensics and Watermarking (IWDW11)


This figure shows returned results in the top-5 images of our and exhaustive methods given the watermarked query images shown in the top row. We do not show the top-1 images since the query images are returned at the top-1 images in all the cases. Sub-captions from (a) to (p) represent image attack used to create the corresponding images. Top-6 to top-10 images are available at the supplementary report.

Abstract

We present a novel, Image Retrieval based Image Watermark (IRIW) framework to identify copyright-violated images in both efficient and accurate manner for large-scale image databases. We first perform SIFT-based image retrieval to identify similar images given a query image and store them as an output list. Then we extract watermark patterns and check watermark similarity only for images stored in the list. As a final step, we re-rank images by considering various information available between each image in the list and the query image and by utilizing information even among images in the list. Also, in order to reduce any negative impacts on image retrieval by embedding watermark patterns on images, we propose to use a SIFT-aware image watermark detection method. Compared with the exhaustive method that checks all the images stored in an image database that consists of 10 K images, our method achieves more than two orders of magnitude performance improvement. More importantly, by identifying similar images given a query image and focusing on checking watermark similarities among those similar images, we are able to reduce false positive and false negative cases by a factor of up to two over the exhaustive method.

Contents

Paper: IRIW Paper (PDF, 6.4MB) , Supplementary Report PDF, 9.3MB)


@inproceedings{Jun11,
author = {Jong Yun Jun, Kunho Kim, Jae-Pil Heo, Sung-eui Yoon},
title = {{IRIW}: Image Retrieval based Image Watermarking for Large-Scale Image Databases},
booktitle = {IWDW11: 10th International Workshop on Digital-forensics and Watermarking},
year = {2011}
}

Dept. of Computer Science
KAIST
373-1 Guseong-dong, Yuseong-gu, Daejeon, 305-701
South Korea
sglabkaist dot gmail dot com