3D Mesh Watermark Attack Based On Adaptive Noise

As an important branch of three-dimensional digital watermarking algorithm, the purpose of three-dimensional digital watermarking removal algorithm is to identify weaknesses and vulnerable reasons of the three-dimensional digital watermarking system , then be improved. This study as experienced by traditional password encryption - decryption improve and promote each other ; the three-dimensional digital watermark attack algorithm and three-dimensional digital watermarking algorithm are complementary and mutually promoting. In this paper, we study the three-dimensional mesh digital watermarking removal algorithm.Three-dimensional mesh watermarking removal algorithm attacks the three-dimensional model which have been added watermark on.According to the different attack methods, there is direct removal of watermark, the watermark synchronization damage, etc., which make the watermark not to detect by watermarking system. Whatever the mode of attack methods are, they have to face the same problem is that should not bring about discernible deformation to the model and should not affect the model's use. The watermark strength and attack strength are constraint by each other and limited by the normal availability of the model. By the first step 1 - Neighborhood adjustment to model, this paper weak and limit the watermark strength to 1 - Neighborhood of vertex. And then attacks is performe under the intensity of 1 - Neighborhood. So after the 1 - Neighborhood adjustment, the attack methods is only considered rather than considering the strength and the methods of the attack both.In this paper, the major improvements is on two attack methods, smoothing attack and noise attacks. Smoothing attacks use Laplacian smoothing and averaging normal smoothing method, in view of the details of different geometric characteristics of the model. The different geometric characteristics of the model of of is represented by the curvature. The low curvature region are attacked by Laplacian smoothing method; the high curvature region is manifestation of the model's details, and we use the improved averaging method smoothing to attack the regions, which show good results in maintaining the details of the model.In various regions of model we selecte sampling points, and with sets of binary form (curvature, noise level) record of such sampling point. With the recorded binary form we use a curve fitting interpolation to fit the adaptive noise curve,and then use this curve to calculate the noise strength of the remaining points. After that we select the direction of noise in a orientation sets and add noise to all vertices of the model.The experimental results proved that there is a better average performance without a clear deformation of medal compared with other three-dimensional mesh of digital watermarking attack algorithms, by the adoption of three-dimensional mesh digital watermarking attack algorithm.