Reconstructing Objects from Noisy Images at Low Resolution

Research output: Chapter in Book/Report/Conference proceedingArticle in proceedingsResearchpeer-review

Helene Svane, Aasa Feragen

We study the problem of reconstructing small objects from their low-resolution images, by modelling them as r-regular objects. Previous work shows how the boundary constraints imposed by r-regularity allows bounds on estimation error for noise-free images. In order to utilize this for noisy images, this paper presents a graph-based framework for reconstructing noise-free images from noisy ones. We provide an optimal, but potentially computationally demanding algorithm, as well as a greedy heuristic for reconstructing noise-free images of r-regular objects from images with noise.

Original languageEnglish
Title of host publicationGraph-Based Representations in Pattern Recognition - 12th IAPR-TC-15 International Workshop, GbRPR 2019, Proceedings
EditorsDonatello Conte, Jean-Yves Ramel, Pasquale Foggia
PublisherSpringer
Publication date2019
Pages204-214
ISBN (Print)9783030200800
DOIs
Publication statusPublished - 2019
Event12th IAPR-TC15 Workshop on Graph-Based Representations in Pattern Recognition, GbRPR 2019 - Tours, France
Duration: 19 Jun 201921 Jun 2019

Conference

Conference12th IAPR-TC15 Workshop on Graph-Based Representations in Pattern Recognition, GbRPR 2019
LandFrance
ByTours
Periode19/06/201921/06/2019
SeriesLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume11510 LNCS
ISSN0302-9743

    Research areas

  • Object reconstruction, r-regularity

ID: 227334305