Intrinsic/extrinsic embedding for functional remeshing of 3D shapes

3D acquisition pipeline delivers 3D digital models accurately representing real-world objects, improving the geometric accuracy and realism of virtual reconstructions. However, even after intensive clean-up, the captured models fall short of many of the requirements imposed by the downstream application, such as video-games, virtual reality, digital movies, etc. Often, the captured 3D model can only serve as a starting point for a cascade of subsequent phases, by either digital artists or geometry processing algorithms, such as a complete remeshing (or retopology), surface parameterization, and skinning for animation. In contrast, we propose a novel remeshing-by-matching approach, where we automatically combine the accurate 3D geometry of the captured model with the tessellation of a target pre-existing template which already satisfies all the professional requirements. At the core of this process, there is a matching strategy based on the functional mapping framework. To this end, we introduce a new set of basis functions designed for this context: termed Coordinates Manifold Harmonics (CMH). We evaluate this strategy (quantitatively and qualitatively) over models of different classes, obtaining a favourable comparison with existing methods.