Normal compression based on clustering and relative indexing

While the compression of topology and geometry has been explored significantly, the same issue for normals has not yet been studied as much as it deserves. Presented in this paper is a better approach than existing ones to compress the normals of a mesh model. The proposed scheme uses relative indexing to refer to the normals from each face definition. Besides, this scheme uses the concept of clustering of model normals so that the distribution of normals is considered.