A robust 3D point cloud watermarking method based on the graph Fourier transform

Many modern applications make use of 3D modeling and / or reconstruction of complex objects, such as historical monuments and entire urban centers. One of the most common representations of these 3D models is by point clouds, which are a dense set of points irregularly organized in a 3D coordinate system. Usually, the acquisition methods are highly expensive due to the necessary equipment and size of these models. This factor motivates the proposal of watermarking techniques to guarantee the copyright protection as well as to detect illegal copies. This paper presents a non-blind watermarking method for 3D point clouds. The method is based on the graph Fourier transform, a recently introduced signal processing tool which has been applied to signals lying over arbitrarily irregular domains. Unlike other published works regarding point cloud watermarking, instead of inserting the bit sequence in the models’ spatial coordinates, in this work the bits are embedded in the color information attributed in each point of a cloud. Simulation results show high imperceptibility and robustness against several attacks, such as affine transformations, reordering, noise addition and cropping.