Robust Zero-watermarking Scheme Using Local Invariant Keypoints

This paper presents a novel zero-watermarking scheme, which is robust to geometric distortions and common signal processes. Zero-watemarking technique is different from traditional digital image watermarking, which constructs watermark from its host image, instead of watermark inserting. We construct watermark from low-frequency coefficients in discrete wavelet transform (DWT) domain, which is robust to signal processes. Before DWT, we performs log-polar mapping (LPM) in the host image for being against scaling and cropping. For translation and rotation invariance, we apply Scale Invariant Feature Transform (SIFT) feature descriptors to locate two local invariant keypoints. One of them is regarded as the original of LPM and both of them are as reference points in correcting rotation distortion. Extensive experiments show that our scheme is effective and outperforms the previous watermarking schemes in resisting signal process, aspect ratio change, scaling and shearing. Key-Words: Zero-watermarking; Discrete wavelet transform; Scale invariant feature transform; Log-polar mapping; geometrical distortion.