A variational approach for physically based image interpolation across boundaries
In this contribution we present an optimal control approach for physics-based optical flow estimation and image interpolation. The aim of the developed process is to identify appropriate boundary data of an underlying physical model describing the transport field, which reason the movement of an ini...
Gespeichert in:
| 1. Verfasser: | |
|---|---|
| Dokumenttyp: | Kapitel/Artikel Konferenzschrift |
| Sprache: | Englisch |
| Veröffentlicht: |
2015
|
| In: |
Multiple Shooting and Time Domain Decomposition Methods
Year: 2015, Pages: 315-354 |
| Schlagworte: | |
| Online-Zugang: | Verlag, Volltext: https://link.springer.com/chapter/10.1007/978-3-319-23321-5_13 Verlag, Volltext: https://link.springer.com/content/pdf/10.1007%2F978-3-319-23321-5_13.pdf |
| Verfasserangaben: | Matthias Klinger |
| Zusammenfassung: | In this contribution we present an optimal control approach for physics-based optical flow estimation and image interpolation. The aim of the developed process is to identify appropriate boundary data of an underlying physical model describing the transport field, which reason the movement of an initial brightness distribution. Thereby, the flow field as solution of the time-dependent non-linear Navier-Stokes equations is coupled to a transport dominant convection-diffusion equation describing the brightness intensity. Thus, we have to deal with a weakly coupled PDE system as state equation of a PDE constrained optimisation problem. The data is given in form of consecutive images, with a sparse temporal resolution, representing the brightness distribution at different time points. We will present the mathematical theory of the resulting optimisation problem, which is based on a Robin-type boundary control. We describe the numerical solution process and present by means of synthetical test cases the functionality of the method. Finally we discuss the application of multiple shooting techniques for the considered problem, since we observed that the employed Newton-type method is very sensitive with respect to the chosen initial value. |
|---|---|
| Beschreibung: | Gesehen am 06.11.2017 |
| Beschreibung: | Online Resource |
| ISBN: | 9783319233215 |