Triangular meshes: A solution to resist to geometric distortions based watermark-removal softwares

We present in this paper a compensation technique allowing to retrieve a watermark in an image attacked by random local geometric distortions, with the help of the original image, or its edges. For this purpose, we use a flexible triangular 2-D mesh, in order to compensate parts of geometric distortions that could be introduced by softwares such as Stirmark [12, 11]. In order to illustrate the usefulness of the compensation on attacked images, we use two watermarking techniques: a classical spread spectrum algorithm proposed by Cox et al. [3] (the watermark is embedded into a selected set of DCT coefficients) and a wavelet-based watermarking technique [4].