Difference expansion based reversible watermarking for 3D mesh models

Abstract

As applications involving 3D data are increasing, 3D watermarking algorithm is drawing attention of the researchers. However, classical watermarking algorithms often make distortion to the original model. In some fields like medical imagery or military information, any change of the original model is not allowed. For those fields, reversible watermarking which can recover the original model is required. The performance of reversible watermarking of 2D image is at high level, but there is a room to improve 3D watermarking technology. In this thesis, we propose a reversible watermarking algorithm for 3D meshes based on difference expansion. We embed watermark in the model by expanding the difference between the vertices and their adjacent vertices. We first traverse the mesh model to choose the vertices into which the watermark bits will be embedded. We use all the adjacent vertices of a vertex for the expansion, so that the capacity is improved while keeping low distortion. Compared to state-of-the-art reversible 3D watermarking algorithms, our algorithm shows high capacity and low distortion.