Abstract: In recent years, existing 3D digital watermarking algorithm focuses more on the robustness against adverse attacks rather than the visual quality. This paper proposes a new method based on Discrete Wavelet Transform（DWT）, which in large measure improved the visual quality of the embedded 3D content. This algorithm selects the traditional 3D triangular mesh and begins by calculating the mesh saliency region of the 3D content, in the meantime, conducting the process of the Principal Component Analysis（PCA） to original 3D content, which obtaines the powerful robustness to the affine attacks such as translation, scaling and rotation attacks. Then it conducts the discrete wavelet transform to the spherical coordinate of the processed 3D content, embedding the watermark information into both low coefficients and high coefficients. The final watermarked 3D content can be obtained through the inverse discrete wavelet transform of the modified coefficients. The experimental results show that this algorithm not only has the strong robustness to possible attacks but also keeps the excellent visual quality of the 3D content.

Key words: 3D watermarking, Discrete Wavelet Transform（DWT）, Principal Component Analysis（PCA）, mesh saliency, affine attacks

摘要： 大部分三维数字水印算法更多地关注水印的鲁棒性而忽视了三维模型的视觉效果。为了解决这一问题，提出了一种基于小波变换的三维网格数字水印新算法，该算法以典型的三角网格模型作为水印载体。提取三维模型中的显著区域，并将三维模型进行旋转归一化处理，以此保证对三维模型平移、缩放或旋转攻击都有很强的鲁棒性;将三维模型转化到球面坐标下并进行小波变换，在低频和高频系数下嵌入水印；经过小波逆变换得到嵌入水印后的三维模型，实验结果表明提出的算法既对多类攻击方式具有鲁棒性，又保持了三维模型的视觉效果。

关键词: 三维水印, 离散小波变换（DWT）, 主成分分析（PCA）, 网格显著区域, 仿射攻击