Cage‐Based Dynamic Mesh Compression

Cage‐based deformation is a convenient way to create 3D animations of high‐resolution meshes by means of moving the vertices of a low‐resolution control mesh, the cage. The resulting animation is a dynamic mesh, that is, a sequence of meshes with shared connectivity, which is controlled entirely by the corresponding dynamic cage. In this paper, we propose to invert this process and to approximate a given mesh sequence with a cage‐based animation of the average mesh of that sequence. This process effectively decomposes the data into a static shape, represented by the average mesh, and the motion, given by a dynamic cage sequence. One key application of this representation is data compression. We use mean value coordinates for binding the mesh to the cage and publicly available state‐of‐the‐art algorithms for computing the average mesh and for compressing static and dynamic meshes. Apart from the general idea of inverse cage‐based animation, we describe a novel approach for automatically creating a motion‐aware cage for the average mesh as well as a method for optimizing the latter. Compared to direct compression of the given dynamic mesh, our overall pipeline turns out to give better results at high compression rates, which are the most relevant in practice. Moreover, the decomposition into shape and motion enables the user to modify the input sequence by editing the average mesh and propagating the changes to the whole sequence using the dynamic cage.