Double Zero-Watermarks Scheme Utilizing Scale Invariant Feature Transform and Log-Polar Mapping

This paper presents a novel watermarking scheme that constructs two zero-watermarks from its host image. One is robust to signal process and central cropping, which is constructed from low-frequency coefficients in discrete wavelet transform (DWT) domain of its host image; the other is robust to general geometric distortions as well as signal process, which is constructed from DWT coefficients of log-polar mapping (LPM) of its host image. During the second watermark generating, we select two scale invariant feature transform (SIFT) feature descriptors to locate two invariant points, one of which is regarded as the origin of LPM and both are as reference points in correcting rotation distortion. Experiments show that our scheme is effective and outperforms the previous watermarking schemes in resisting signal process, aspect ratio change, scaling and shearing.