A Non-blind Robust Watermarking Scheme for 3D Models in Spatial Domain

Abstract

Improving robustness against various attacks is the main objective for 3D spatial watermarking scheme. Aimed at the goal, a novel spatial robust watermarking scheme is proposed. The Cartesian coordinates of 3D model vertices are converted into spherical coordinates. All vertices are partitioned and sorted by angles, and then vertex sets for watermark embedding are built. The vertex norms in one vertex set are modified to embed a watermark bit and the same watermark bit is embedded repeatedly into various vertex sets. The classification and modification of vertices are separated to avoid de-synchronization problem. In watermark detection, a watermark bit is extracted according to the number of vertices satisfied with specified conditions by comparing the detected model and the original model. Experiments show that the watermarking scheme is robust against attacks such as translation, rotation, uniform scaling, vertex reordering, noise, simplification, cropping, quantization, smoothing, subdivision and their combined attacks.