Optimizing Geometry-aware Pants Decomposition

Abstract

We present a computational framework to optimize the pants decomposition of surfaces with non-trivial topology. A pants decomposition segments a surface into a set of sub-patches each of which is genus-0 with 3 boundaries. A given surface usually admits infinitely-many pants decompositions that are topologically inequivalent. Given some pre-determined geometric criteria, our algorithm enumerates different classes of pants decompositions and search for the optimal one. The proposed framework is general, and can be used to generate different suitable segmentations according to different applications. We also generalize our algorithm for consistent decomposition of multiple surfaces. This algorithm can be used in constructing compatible cross-surface mapping, and facilitate many computer graphics tasks.