Correspondence estimation for non-rigid point clouds with automatic part discovery

Hao Guo, Dehai Zhu, Philippos Mordohai

Research output: Contribution to journalArticlepeer-review

9 Scopus citations

Abstract

We propose an approach for estimating non-rigid correspondences between two shapes that can handle articulation and deformation of the surfaces to be matched. It operates on open or closed surfaces represented by point clouds and, therefore, it can be applied on other representations that can be converted into point clouds. Our method is capable of automatically discovering the articulated parts of the surface without requiring knowledge of the topology or the number of rigid parts. Processing begins by estimating potential sparse correspondences between the source and the target surface. These are used to align the largest corresponding parts of the two surfaces. Fragments of the surface that are not consistent with this alignment generate part hypotheses on which the algorithm is applied recursively. We present qualitative and quantitative results on four datasets comprising open and closed surfaces.

Original languageEnglish
Pages (from-to)1511-1524
Number of pages14
JournalVisual Computer
Volume32
Issue number12
DOIs
StatePublished - 1 Dec 2016

Keywords

  • Non-rigid correspondence
  • Point clouds
  • Shape matching

Fingerprint

Dive into the research topics of 'Correspondence estimation for non-rigid point clouds with automatic part discovery'. Together they form a unique fingerprint.

Cite this