A Video Dual Watermarking Algorithm Against Geometric Attack Based on Integer Wavelet and SIFT

Abstract

In this paper, a video dual watermarking algorithm is presented which is based on integer wavelet and SIFT (scale invariant feature transform). Firstly, the maximum embedding intensity is acquired which is based on the visual threshold of video content by studying human visual masking model of three-dimensional motion feature in video sequence deeply, and using multiple motion characteristics. Secondly, in order to enhance the ability of resisting geometric attacks and non-geometric attacks, a video frame is divided into low frequency and medium-high frequency by integer wavelet transform, respectively in different watermarking algorithm. For medium-high frequency sub-band coefficients, an adaptive watermarking algorithm is proposed, which is based on video motion information. Then utilizing the stability of low frequency sub-band coefficients histogram under some geometric attacks such as rotation, scaling, and so on, the watermark can be embedded in its adjoining coefficients. And Finally, SIFT which has the scale invariance and direction invariance can be used as the trigger to judge whether the video is subjected to the geometric attacks. SIFT is used to correct video which is subjected to geometric attacks and use the low frequency watermarking extraction algorithm to get the water marking signals. For non-geometric attacks, the medium-high frequency watermarking extraction algorithm is used to get water marking signals. The experimental results proved that our method can more effectively resist the common attacks especially the geometrical attacks and consequently achieve higher robustness.