Research On Watermarking Algorithm For Video Copyright Protection And Content Authentication

With the rapid development of the Internet and mobile communication, multi-media devicessuch as mobile phone, ipad and so on, becomes more and more popular. Thus promote thedevelopment of multi-media industries including video-on-demand, video sharing website andIPTV. Along with this trend, the copyright protection and content security of the multi-media arebecoming an urgent problem. Since video watermark is an efficient method to protect copyrightas well as prevent malicious tampering, it is of great importance to study on it. On the analysis ofthe existing video watermarking algorithm, two algorithms which are suitable for H.264standardare proposed in this paper: a shot-segmentation-based video watermarking algorithm usingmotion vector which is used for copyright protection and a reversible semi-fragile watermarkingalgorithm based on the spatiotemporal analysis which is used for content authentication. Themain work is as follows:(1) To solve the copyright problem of the multi-media, a shot-segmentation-based videowatermarking algorithm using motion vector is proposed. It adopts motion-vector-based shotsegmentation to divide the video into different shots. The accuracy of the shot detection isimproved by this method. Then extract the key frame from each shot as the host frame. Itadaptively selects blocks with large JND (just noticeable difference) in the key frames as the hostblocks and watermark is embedded in the motion vector. The experimental results show that theproposed scheme has a good performance in shot detection, key frame extraction and can findthe appropriate. It has a good invisibility, real-time performance and good performance inresisting Gaussian noise, frame dropping, frame swapping, etc.(2) To verify the video content and detect malicious attacks, a reversible semi-fragilewatermarking algorithm based on the spatiotemporal analysis which is suitable for H.264/AVCvideo is proposed. Firstly, the sum of the predicted and residual energy relationship between twoblocks (or group blocks) in I-frame is utilized to generate content-based feature code. In order toimprove the robustness of the authentication code, the chaos sequence is used to eliminate thecorrelation between energy blocks. Moreover, a spatiotemporal analysis is exploited to guaranteea minimum impact on perceptual quality and bit-rate increment. The watermark is embedded intothe last nonzero coefficient of the selected44sub-blocks. Since it is embedded after the entropy decoding, it doesn’t need motion compensation thus there is no drift error. The experimentalresults show that the propose method excellent in terms of perceptual quality and bit-rate, andhas a good performance in detecting malicious attacks. The original video can be recovered if theauthentication is successful and the tampered area of I frame can be effectively located if theauthentication fails.