Improved autocorrelation function based watermarking robust to geometric distortion

Abstract

Geometric distortion is regarded as one of the strongest attacks in the digital watermarking world. Even slight rotation, scaling, and cropping can disable most watermarking algorithms. Although a few watermarking schemes that can handle the geometric distortion have been introduced, they still have a lot of problems. In this thesis, an improved ACF (autocorrelation function) based watermarking using informed embedding mechanism is proposed. The ACF based watermarking is thought of as one of the most effective watermarking schemes that resist geometric attacks. In this watermarking scheme, the autocorrelation peaks of the watermark play an important role for the geometric attack estimation. If the peaks are removed, the watermark detection fails even if the watermark still remains in the image. The problem is that the peaks are not strong enough. The proposed scheme improves the peak strength by using the information of an original image sophisticatedly during the watermark embedding. By making the image contain highly autocorrelated signal, stronger autocorrelation peaks are obtained in the detection process. Experimental results show that the proposed scheme yields stronger autocorrelation peaks than conventional ACF based watermarking. The proposed scheme provides improved watermark detection performance against geometric and combined geometric-removal attacks.

We also propose an ACF based watermarking in the discrete wavelet transform domain. Generally, it is known that watermarking methods in the frequency domain outperform spatial domain methods. The conventional ACF based watermarking embeds watermark in the spatial domain due to its special detection mechanism. In this thesis, we show that autocorrelation peaks are also extracted by embedding a periodic watermark into the wavelet subbands of an image. Using this property, we present the embedding and detection method of the ACF based watermarking in the DWT domain. The characteristics of the peaks...