Robust Key-point based Data Hiding

Anindya Sarkar, Lakshmanan Nataraj and B. S. Manjunath,
Department of Electrical and Computer Engineering,
University of California, Santa Barbara
anindya [at] ece.ucsb.edu

Abstract

We propose a robust data-hiding method which can survive a host of global, geometric and image editing attacks. Localized redundant embedding around multiple key-point centered regions are used to provide robustness to local attacks and geometrical transformations. The primary contributions of the work include novel methods for introducing synchronization information that can be recovered at the decoder without any side information. This synchronization information is then used to identify accurately the geometrical transformation. We show through experiments that our approach is robust to many standard image processing attacks, including JPEG type compression. Quantization index modulation based embedding and repeat accumulate code based error correction are used to obtain a good trade-off between hiding rate and error resilience. Our hiding scheme is robust against rotation angles within (−45 degrees, 45 degres) and scaling factors within (0.5,2). Detailed experimental results are provided.
[PDF] [BibTex]
Anindya Sarkar, Lakshmanan Nataraj and B. S. Manjunath,
VRL, ECE, University of California, Santa Barbara, Sep. 2009.
Node ID: 527 , DB ID: 334 , Lab: VRL , Target: Technical Report
Subject: [Digital Watermarking and Data Hiding] « Look up more