Research On Algorithms Of Video Watermark Oriented To AVS

As a method of digital multimedia copyright protection, digital watermark technology has developed rapidly in recent years, it has first been studied for still images, with the popularization of video products, video watermark becomes a hotspot and difficulty of digital watermark, has wide application prospect. Under the background of large-scale development of the multimedia industry, the AVS (Audio Video coding Standard) with independent property is proposed by our country. As a foundational standard of digital audio and video system, AVS is gradually recognized in China, and also attracted great attention in the international community. The study on the AVS video watermark technology is of great significance and practical value. This paper presents two AVS-oriented digital video watermark methods.This paper proposes an AVS video watermark algorithm based on gray reational analysis and grey prediction . The algorithm embedds watermark during the process of AVS video coding, detailed as follows: Firstly, extract the watermark from the original gray image with gray prediction, compress the embedded information and convert the watermark into binary sequences; Secondly, calculate the grey correlation degree between the host video blocks, select the blocks with smallest correlation degree, then perform uniform spectrum process on the embedded region; Finally, adjust the parity of the quantization coefficients to embed watermark. When extracting the watermark, determine the extracting location in accordance with the key, extract the binary watermark and then recover the gray image watermark. The experiment shows that the proposed method has big embedding capacity, and is robust to common attacks.This paper proposes an AVS video watermark algorithm based on low dimensional feature. Firstly, the algorithm carries on dimensionality reduction to the video frame, completes video division according to the similarity of neighboring frames's low manifold; Then dynamic watermark based on the low dimension characteristics of the I frame sequences is produced online; The HVS (Human Visual System) model is improved by movement information according to the video's low dimension characteristics and the watermark embedding intensity is adaptively controled by the improved HVS model. The experiment shows the algorithm balances the contradiction between the transparency and the robust effectively, and robust to attacks such as compression, noise, filter, cutting, average of frames, frame regrouping. Finally, the summary of this paper is given and the future direction of the research is presented.