Video Watermarking Resistant To Geometric Attacks

With the development of digital watermarking techniques, digital video watermarking has become an effective way to protect the copyright of multimedia video. However, the requirement of robustness is not only to common video distortion, but also to a special attack type, namely, geometric attack. Unlike common video watermarking distortion, geometric attack influences not only the watermark but also the related additional information, such as the embedding position and orientation. In this paper, we will research digital video watermarking resistant to geometric attack.Two kinds of digital video watermarking schemes resistant to geometric attacks are proposed. For the first approach, one-level wavelet transform along time axis is performed on a video shot, and then the binary watermark is scrambled and embedded into the geometric invariant regions extracted from the low-pass frames. The embedding parameter is selected based on the content, and the watermark extraction can be achieved without original video.For another approach, based on embedding sinusoidal signals pattern into images to form some peaks in frequency domain, sinusoidal signals pattern as watermark is embedded individually into the selected sub-bands of 3D complex dual-tree wavelet transform domain(3D-CWT). The HVS masking is used to decide the watermark embedding strength. Then, in detection process, we make use of the peaks response of watermark to estimate and rectify the attacks suffered. Furthermore, confirmation of watermark existence is achieved by detecting the prominent peak in frequency domain.The experimental results demonstrate that the proposed video watermarking schemes not only have good imperceptibility but also robust against typical common video distortions and geometric attacks.