Watermarking 3D Triangular Mesh Models Using Intelligent Vertex Selection

Watermarking provides a mechanism for copyright protection or ownership assertion of digital media by embedding information in the data. In this paper, a blind robust 3D triangular mesh watermarking approach is presented using one of Computational Intelligent (CI) techniques named neural network; the role of this neural network is to select the best vertices that can be used as watermark carrier. This watermark position has to guarantee minimum distortion of 3D model and maximum robustness for watermark bits extraction. First, we select the best position of watermark carrier vertices using smoothing feature clustering. This clustering stage is performed using one of unsupervised neural network types which is a Self Organizing Maps (SOM). Then, watermark bits stream are embedded in the selected marked vertices using local statistical measures such as; mean and standard deviation. Experimental results show that our watermarking algorithm is robust since watermarks can be extracted without mesh alignment or re-meshing under a variety of attacks, including noise addition, cropping, smoothing filtering, rotation, translation, and scale. Our work had been compared with other work of blind 3D watermarking models and proves its efficiency in terms of both robustness and imperceptibility.