Adaptive Unified Data Embedding and Scrambling for Three-Dimensional Mesh Models

Traditionally, unified data embedding and scrambling techniques have been designed for grayscale images, which cannot be applied directly to a three-dimensional (3D) mesh. Recently, the universal use of 3D technology inspired us to innovate in this field. In this paper, an adaptive unified data embedding and scrambling technique for 3D mesh models (3D-AUES) is proposed, which can embed external data and scramble 3D mesh simultaneously. First, a vertex coordinate prediction method called cross prediction is adopted to accurately predict half of the vertices from the other half. The predicted vertices are used to embed external data. We further increase the embedding rate by bit replacement embedding. Then, to improve security, we propose an adaptive threshold to select vertices for embedding. To ensure lossless scrambling, the thresholds and prediction errors are embedded as side information with secret information into the vertices. By adopting an adaptive threshold and multilayer embedding, scalable scrambling quality can be achieved. On the decoder side, with the help of losslessly embedded side information, external data can be successfully extracted, and the original mesh can be restored to predetermined distortion levels, from lossless recovery to partial recovery. Experiments show that 3D-AUES has a high embedding rate, scalable scrambling quality and scalable recovery quality.