Watermarking raw 3D point clouds preserves geometry and ownership
Geometry-Preserving Blind Watermarking for Raw 3D Point Clouds
Computer Vision and Pattern Recognition
Summary
It is hard to prove who owns 3D point clouds, which are collections of dots representing shapes or scenes, because the points can be rearranged or changed easily. The authors created a method to hide a secret message directly in the 3D points that can be checked later without needing the original data. Their method uses a type of neural network that understands the shape at multiple scales and trains to resist common changes like rotations or resampling. Tests show the hidden message can be found reliably without messing up the shape or making visible changes to the points.
What this means in practice
- •For 3d scanning companies: Embed ownership information invisibly into 3D scans to protect rights without altering scan quality.$Commercial implications: Enables selling of secure 3D scanning products with built-in proof of ownership for clients.
- •For virtual reality developers: Verify the authenticity of detailed 3D environments by recovering embedded ownership data from raw point clouds.
Authors
Rungui Zhou, Chuanzhi Zhou, Ruihuan Wang, Peng-Shuai Wang
Abstract
Raw 3D point clouds are a core geometric representation. Establishing their ownership is challenging because point sets are irregular, unstructured, and frequently altered by resampling and geometric preprocessing. We present a blind watermarking framework that operates directly on xyz coordinates and supports both object-level shapes and scene-scale scans. At verification time, the embedded message is recovered from the observed point cloud alone, without access to the original point cloud, color, normals, or mesh connectivity. The method jointly learns watermark embedding and extraction through a feed-forward octree-based architecture, enabling efficient multi-scale geometric reasoning on large point sets. During training, a stochastic transformation layer exposes the decoder to common geometric perturbations, while progressive pose alignment improves robustness to pose changes. Experiments on object-level and scene-level benchmarks demonstrate reliable message recovery under common geometric processing while maintaining low geometric distortion. Qualitative comparisons further show that the learned perturbations are less visually conspicuous and less spatially structured than those of handcrafted alternatives.