¡@

Former watermarking techniques have been largely developed for application to natural videos. Important applications, including virtual reality, computer games, cartoons, and movies, involve another category of visual data: computer animation. In this study, we developed robust watermarking techniques to protect MPEG-4 2D mesh animation against copyright infringement. Three time-series analysis tools, the discrete cosine transform (DCT), singular spectrum analysis (SSA), and discrete wavelet transform (DWT), are employed to analyze the motion characteristics of mesh animation. The watermark is cast upon the important motion components by employing the spread-spectrum principle. During watermark extraction, a spatial-domain least-squares registration technique is used to restore the distorted mesh. Each watermark bit is then detected by a hard decision with the aid of cryptographically secure keys. We have tested the proposed system against a variety of attacks, including affine transformations, temporal smoothing, spectral enhancement and attenuation, and additive random noise. The time-series analysis tools studied here are evaluated in terms of their robustness to attacks and the required computational complexity.

Keywords: copyright protection, computer animation, discrete cosine transform (DCT), discrete wavelet transform (DWT), MPEG-4, singular spectrum analysis (SSA), time-series analysis, watermarking

Retrieve PDF document (200503_06.pdf)

Received May 11, 2004; revised July 20, 2004; accepted September 1, 2004.
Communicated by Ja-Ling Wu.
^* This work was supported in part by National Science Council, Taiwan, R.O.C., under Contract No. NSC 90-2213-E-027-012. A preliminary version of this paper was presented in Information Security Conference (ISC2004), 2004.