Digital watermarking of polygonal meshes with linear operators of scale functions

Digital watermarking is already used to establish the copyright of graphics, audio and text, and is now increasingly important for the protection of geometric data as well. Watermarking polygonal models in the spectral domain gives protection against similarity transformation, mesh smoothing, and additive random noise attacks. However, drawbacks exist in analyzing the eigenspace of Laplacian matrices. In this paper we generalize an existing spectral decomposition and propose a new spatial watermarking technique based on this generalization. While inserting the watermark, we avoid the cost of finding the eigenvalues and eigenvectors of a Laplacian matrix in spectral decomposition; instead we use linear operators derived from scaling functions that are generated from Chebyshev polynomials. Experimental results show how the cost of inserting and detecting watermarks can be traded off against robustness under attacks like additive random noise and affine transformation.