Schrödinger diffusion for shape analysis with texture

Jose A. Iglesias, Ron Kimmel

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

In recent years, quantities derived from the heat equation have become popular in shape processing and analysis of triangulated surfaces. Such measures are often robust with respect to different kinds of perturbations, including near-isometries, topological noise and partialities. Here, we propose to exploit the semigroup of a Schrödinger operator in order to deal with texture data, while maintaining the desirable properties of the heat kernel. We define a family of Schrödinger diffusion distances analogous to the ones associated to the heat kernels, and show that they are continuous under perturbations of the data. As an application, we introduce a method for retrieval of textured shapes through comparison of Schrödinger diffusion distance histograms with the earth's mover distance, and present some numerical experiments showing superior performance compared to an analogous method that ignores the texture.

Original languageEnglish
Title of host publicationComputer Vision, ECCV 2012 - Workshops and Demonstrations, Proceedings
Pages123-132
Number of pages10
EditionPART 1
DOIs
StatePublished - 2012
EventComputer Vision, ECCV 2012 - Workshops and Demonstrations, Proceedings - Florence, Italy
Duration: 7 Oct 201213 Oct 2012

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
NumberPART 1
Volume7583 LNCS

Conference

ConferenceComputer Vision, ECCV 2012 - Workshops and Demonstrations, Proceedings
Country/TerritoryItaly
CityFlorence
Period7/10/1213/10/12

Keywords

  • Laplace-Beltrami operator
  • Schrödinger operators
  • diffusion distance
  • earth mover's distance
  • textured shape retrieval

All Science Journal Classification (ASJC) codes

  • Theoretical Computer Science
  • General Computer Science

Fingerprint

Dive into the research topics of 'Schrödinger diffusion for shape analysis with texture'. Together they form a unique fingerprint.

Cite this