Research Of Watermarking Based On AVS-M

With the development of image watermarking technology, more and more attention is paid to video watermarking. The special performance requirement and emerge of new attack methods are challenged to the development of video watermarking. AVS-M has many advantages such as good performance and easy patent issue. Watermarking research on AVS-M encoder has good foreground.According to AVS-M standard a new watermarking embedded algorithm is proposed. First, watermarking image is pre-manipulated with chaos. And then edge detecting is used because edge and texture areas are more suitable to embed watermarking, the decoded data of encoder of AVS-M is used to detect edge and decides whether a macroblock is edge-block or not. Finally, edge-macroblock in one P frame with the biggest motion vector magnitude is selected to embed watermarking. Watermarking technology based on motion vector has little cost of calculation and can easy realize real-time requirement.The blocks which motion vector has been modified may cause error to the following blocks during inter-prediction, the error will accumulate without compensation. The influence of modified motion vector to the following blocks is carefully analyzed and compensation methods are given. Many format video sequences are tested, the experimental results show that the watermarking correct ratio is 100%, but it is 60% without the compensation methods.Watermarking embedded methods based on motion vector require synchronous detection, a new detecting way named Maximum-Likelihood sub-sequence vector detection was given, which can resist delete or insert attacks. The experimental results show that the method is efficient.Besides, watermarking based on DCT is researched and an algorithm is also developed.