Geometry Compression of Triangle Meshes using a Reference Shape

Triangle mesh compression is an established area, however, some of its special cases are yet to be investigated. This paper deals with lossy geometry compression of manifold triangle meshes based on the EdgeBreaker algorithm using a reference shape known to both the encoder and the decoder. It is assumed that the shape of the reference object is similar to the shape of the mesh to be encoded. The predictions of vertices positions are done extrinsically, i.e. outside the reference shape, and then orthogonally projected onto its surface. The corrections are encoded by two integer numbers, denoting the layer order and an index of a hexagon in a hexagonal grid generated on the surface of the reference shape centered at the prediction point. The availability of a reference mesh results in a smaller bitrate needed for comparable error when compared to a state of the art static mesh compression algorithm using weighted parallelogram prediction.