Blind Watermarking Technique Robust To Image Compression

Abstract

The security and copyright issue is an emerging topic in multimedia communications. We present an invisible watermarking scheme to embed copyright or ownership information into digital image by altering the selected wavelet transformed coefficients. The watermark is embedded in the transform domain by confining each selected wavelet coefficient to a specified value to embed one bit of information. Watermark detection does not require the existence of the original image nor original watermark. Furthermore, to meet the security requirement, watermark casting and detection agree with a 3DES cipher key of embedding and extracting to prevent hackers from detecting or removing the watermark. Experimental results demonstrate that the watermark is robust to various lossy compression, e.g., those that result from vector quantization (VQ) based compression, DCT based compression (JEPG), and the state-of-the-art wavelet based compression (SPIHT). The scheme could also be applied to content annotation, covert communication channel, access control, content authentication and integrity verification.