Robust Video Watermaking Technology Research Based On DCT

With the rapid development of network technique and multimedia technology,copyright protection of digital viedo becomes a dead-hand problem to be resolved.Digital watermarking technology is an effective means in this field that falls into twocategories: the embedding based on compression domain and original video.Watermarking algorithm combining with the MPEG-4standard is a currentresearching hotspot. The challenges faced by digital video watermarking technologyare invisibility, robustness and high compression performance. At present, there aremany video watermarking algorithm based on MPEG-4proposed at home and abroad.This paper proposes two different robust watermarking algorithms based on discretecosine transform aiming at MPEG-4: an adaptive algorithm based on human visualsystem and a watermarking algorithm based on optical flow.(1) An adaptive video watermarking algorithm based on discrete cosinetransform is presented. With the texture masking features of human visual systemtaken into account, the watermark is embedded by modifying direct-currentcoefficient in discrete cosine transform domain of component of I frame, which hashigh invisibility threshold. The algorithm adaptively adjusts watermarking intensityaccording to the texture features of each block to guarantee both of the capacity andthe reduction of block effects caused by modifying direct-current coefficient.Besides, scrambling encryption technology educes the relativity amongwatermarking data and hugely increases the security of watermarking.(2) This paper proposes a video watermarking algorithm based on image opticalflow. According to the optical flow rate of the points in video images computed byan improved Horn-Schunck approach, the total optical flow of frames is calculated.The frames with high optical flow are selected as the key frames. Then thewatermark is embedded by modifying the middle frequency coefficient in discretecosine transform domain of component U of these key frames. The algorithm adoptsadaptive weight value to increase the accuracy and compensation algorithm todecrease the iteration times and computational complexity and improve the real-timecapability.Simulation experiments are made under the same environment on the base oftheory research. It’s showed that the improved algorithms are workable and have better invisibility and robustness compared to the discrete cosine transformalgorithm of the kind.