Multi-Shape registration with constrained deformations

Research output: Chapter or section in a book/report/conference proceedingChapter in a published conference proceeding

Abstract

Based on a Sub-Riemannian framework, deformation modules provide a way of building large diffeomorphic deformations satisfying a given geometrical structure. This allows to incorporate prior knowledge about object deformations into the model as a means of regularisation [10]. However, diffeomorphic deformations can lead to deceiving results if the deformed object is composed of several shapes which are close to each other but require drastically different deformations. For the related Large Deformation Diffemorphic Metric Mapping, which yields unstructured deformations, this issue was addressed in [2] introducing object boundary constraints. We develop a new registration problem, marrying the two frameworks to allow for different constrained deformations in different coupled shapes.
Original languageEnglish
Title of host publicationGeometric Science of Information - 5th International Conference, GSI 2021, Proceedings
Subtitle of host publicationInternational Conference on Geometric Science of Information
EditorsFrank Nielsen, Frédéric Barbaresco
PublisherSpringer
Pages82-90
Number of pages9
ISBN (Electronic)978-3-030-80209-7
ISBN (Print)978-3-030-80208-0
DOIs
Publication statusPublished - 31 Dec 2021

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume12829 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Bibliographical note

Funding Information:
Rosa Kowalewski is supported by a scholarship from the EPSRC Centre for Doctoral Training in Statistical Applied Mathematics at Bath (SAMBa), under the project EP/S022945/1.

Keywords

  • Large deformations
  • Shape registration
  • Sub-Riemannian geometry

ASJC Scopus subject areas

  • Theoretical Computer Science
  • General Computer Science

Fingerprint

Dive into the research topics of 'Multi-Shape registration with constrained deformations'. Together they form a unique fingerprint.

Cite this