Research On Video Watermarking Algorithm Based On Key Frame

Video watermarking is an important branch in digital watermarking used to protect the videocopyright, which can achieve the goal of identity copyright through embedding specificinformation of copyright in video, and the quality and use value of video are not affected. Withthe rapid development of internet and streaming media technology, video can spread quickly inthe internet of openness and sharing, which raise higher standard for copyright protectionsimultaneously, and become a key research direction.In order to achieve copyright protection and improve video watermarking resistance, twovideo watermarking scheme ware proposed in this paper based on the discrete cosine transform,discrete wavelet transform, combined with the human visual system (HVS) and video features,The specific content of this paper was as follows:(1) Towards the issue of the low ability to resist geometric attacks and high complexity in3D transfer domain video watermarking, this paper presented a blind video watermarkingalgorithm based on key frames and3D wavelet. The key frames were selected based on Eucliddistance of Video frame difference, the key frames were segmented to blocks, and the key blockswere selected based on the video characteristics such as luminance sensitivity and texturesensitivity. Then, the key blocks were reconstructed to three dimension blocks and the scrambledwatermark was adaptively embedded into the low frequency domain of the three dimensionwavelet decomposed coefficients of the blocks. The watermark was retrieved blindly based onJND. Experimental results show that the algorithm improves the watermark real-timeperformance and embedding capacity, and obtained a good imperceptible and robustness to videoattacks.(2) Towards the issue of the lack of relation between video watermarking and video contentin transform domain watermarking, this paper proposed a video zero-watermarking schemethrough region segment for motion object to improve video watermarking robustness andsteganography. Firstly, the key frames were selected based on Euclid distance of video framedifference, the motion object of key frame was selected by a improved frame differencealgorithm, the key frames were segmented to blocks, and the key blocks were selected based onthe centroid of motion object. Then, the key blocks were reconstructed to three dimension block,and the three dimension block was equally segmented to some small three dimension blocks. Thefeature vector was abstracted from the low frequency domain of the discrete cosine transformcoefficients of every block in small three dimension blocks. Lastly, zero-watermarking wasstructured by feature vector. Experimental results show that the algorithm improves the watermarking imperceptibility, and obtain a good robustness to video attacks, especially canresist the compression modes of H.264and rotation.