Geometry-Preserving Blind Watermarking for Raw 3D Point Clouds

2026-09-26

AI Summary

This study addresses the challenges of ownership verification for 3D point clouds arising from their unstructured nature and preprocessing operations. To this end, we propose an end-to-end blind watermarking framework that operates directly on XYZ coordinates. Methodologically, we introduce a pioneering joint learning architecture based on a feed-forward octree, incorporating a random transformation layer and progressive pose alignment to support both object- and scene-level watermark embedding and source-free extraction. Experimental results demonstrate that the proposed method achieves reliable message recovery under common geometric perturbations while maintaining low geometric distortion. Furthermore, the learned watermarks are more imperceptible and natural than those generated by handcrafted approaches, significantly enhancing the robustness and practicality of copyright protection for 3D point clouds.

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.

Problem

• 3D point clouds
• blind watermarking
• ownership verification
• geometry-preserving
• robustness

Innovation

• Blind Watermarking
• 3D Point Clouds
• Octree-based Architecture
• Stochastic Transformation
• Progressive Pose Alignment