Enhanced Watermarking Scheme for 3D Mesh Models

Abstract

In this paper we present a new non-blind watermarking scheme for 3D graphical o b j e c t s (meshes). Non-blind watermarking scheme is known to be more secure than blind ones, since the original and watermarked models are needed for extraction. In our scheme we use the model's prominent feature points to divide the model into separate segments, these feature segments are then projected from 3D representation to the 3 main 2D-Planes. The watermark embedding is done in frequency domain of these projections. The experimental results showed the robustness of this scheme against various mesh attacks (mesh simplification, subdivision, smoothing, cropping, etc). T his scheme also allows quite large payload to embed. The results of the proposed scheme showed average of 50 percent improvement in robustness against geometry attacks, and up to 70 percent against connectivity attacks.

NonBlind Robust 3D Object Watermarking Scheme

“…The baseline evaluation results of the proposed scheme compared to another algorithms [8,13] on Table (2). The results show that we provide much larger payload and RMS with a high redundancy of watermark embedding on the 3D models.…”

“…Table (4) shows a comparison between the proposed algorithm and the previous algorithms [8,9,13] using Bit Error Rate (BER). The results evaluated at different attacks on Venus model.…”

“…The amount of distortion introduced by a watermarking technique is one of the evaluation factors, The RMS measure is not similar, which means that ( ) ( ) the MRMS can be defined as follows [12,13]: ( ) * ( ) ( )+(4) MRMS is more accurate than simple vertex-tovertex measures. BER is used for measuring the robustness of the watermark after the model being attacked which defined as: BER = Errors / Total number of bits.…”

3D model watermarking algorithm robust to geometric attacks

“…Compared with spatial methods, frequency domain methods [15][16][17][18][19][20][21][22] can easily exploit the human visual system to improve the transparency and robustness of the watermark. The watermarking information is embedded in the coefficients of the transform domain, which could be a wavelet transform [15,16], Laplace transforms [17,18], or other transforms [19][20][21][22]. Tamane et al [15] embedded the scrambled watermark in the discrete cosine transform (DCT) coefficients after applying three levels of the Haar transform to the 3D model.…”