Log-scaling magnitude modulated watermarking scheme

A real-time watermarking scheme with high robustness and security has been proposed based on modulating the log-scaling magnitudes of DCT coefficients, which is most suitable for JPEG images and MPEG streams. The watermark bit is encoded as the sign of the difference between the individual log-scaling magnitude of a group-region and the average one of all group-regions. The log-scaling magnitude can be modulated by modifying the low and middle frequency DCT coefficients imperceptibly. The robustness of scheme is not only dependent on those largest coefficients, but also on the other coefficients with the same proportion. It can embed 512 bits into an image with a size of 512×512, which can satisfy the payload requirement of most video watermarking applications. Moreover, the watermark embedding process only requires one-sixth of the time consumed during normal playing of video, and the watermark detection only requires one-twelfth of that, which can meet the real-time requirements of most video watermarking applications. Furthermore, the experimental results show that the presented scheme is transparent and robust to significant valumetric distortions (including additive noise, low-pass filtering, lossy compression and valumetric scaling) and a part of geometric distortions. It performs much better than the EMW algorithm in resisting all kinds of distortions except Gaussian noise with a larger deviation.

Log-scaling magnitude modulated watermarking scheme

A real-time watermarking scheme with high robustness and security has been proposed based on modulating the log-scaling magnitudes of DCT coefficients, which is most suitable for JPEG images and MPEG streams. The watermark bit is encoded as the sign of the difference between the individual log-scaling magnitude of a group-region and the average one of all group-regions. The log-scaling magnitude can be modulated by modifying the low and middle frequency DCT coefficients imperceptibly. The robustness of scheme is not only dependent on those largest coefficients, but also on the other coefficients with the same proportion. It can embed 512 bits into an image with a size of 512×512, which can satisfy the payload requirement of most video watermarking applications. Moreover, the watermark embedding process only requires one-sixth of the time consumed during normal playing of video, and the watermark detection only requires one-twelfth of that, which can meet the real-time requirements of most video watermarking applications. Furthermore, the experimental results show that the presented scheme is transparent and robust to significant valumetric distortions (including additive noise, low-pass filtering, lossy compression and valumetric scaling) and a part of geometric distortions. It performs much better than the EMW algorithm in resisting all kinds of distortions except Gaussian noise with a larger deviation. Supported by the National Natural Science Foundation of China (Grant No. 60502024), and the Natural Science Foundation of Hubei Province (Grant No. 2005ABA267)