Forskning ved Københavns Universitet - Københavns Universitet

Forside

A total variation motion adaptive deinterlacing scheme

Publikation: Bidrag til bog/antologi/rapportKonferencebidrag i proceedingsForskningfagfællebedømt

  • Sune Høgild Keller
  • Francois Lauze
  • Mads Nielsen
We propose a new way of deinterlacing using a total variation scheme. Starting by the Bayesian inference formulation of total variation we do MAP by rewriting the problem into PDEs that can be solved by simple numerical schemes. Normally deinterlacing schemes are developed ad hoc with online hardware implementation directly at eye, sometimes with some frequency analysis as only theoretical base. Our belief is that mathematically well based image models are needed to do optimal deinterlacing and by our work presented here, we hope to prove it. Comparing the output of our scheme with those of ten known deinterlacing schemes shows very promising results.
OriginalsprogEngelsk
TitelScale Space and PDE Methods in Computer Vision : 5th International Conference, Scale-Space 2005, Hofgeismar, Germany, April 7-9, 2005. Proceedings
RedaktørerRon Kimmel, Nir A. Sochen, Joachim Weickert
ForlagSpringer
Publikationsdato2005
Sider408-418
ISBN (Trykt)978-3-540-25547-5
ISBN (Elektronisk)978-3-540-32012-8
DOI
StatusUdgivet - 2005
Eksternt udgivetJa
Begivenhed5th International Conference on Scale Space and PDE Methods in Computer Vision - Hofgeismar, Tyskland
Varighed: 7 apr. 20059 apr. 2005
Konferencens nummer: 5

Konference

Konference5th International Conference on Scale Space and PDE Methods in Computer Vision
Nummer5
LandTyskland
ByHofgeismar
Periode07/04/200509/04/2005
NavnLecture notes in computer science
Vol/bind3459/2005
ISSN0302-9743

ID: 4941739