A Watermarking Method for 3D Models Based on Feature Vertex Localization

This paper presents a new blind watermarking scheme for 3-D point cloud models by using feature vertices to locate the embedded watermark. The vertices with larger mean curvature are considered to be feature vertices carrying watermarking information; non-feature vertices are used to build a new coordinate system in which the 3-D point cloud model is classified into bins containing several longitudes. Each bit of the watermarking information is repeatedly embedded into a bin by altering azimuth angles of feature vertices lying in the bin. The non-feature vertices are not exploited to embed the watermarking information, which enables the built coordinate system to avoid the impacts of the embedding of watermark. The choice of the embedding position and redundant embedding provide an optimal balance between imperceptibility and robustness of watermark. Simulation results testify that our watermarking method for 3-D point cloud models has desired robustness against common attack and geometric attack while demonstrating good transparency.