A Lagged Diffusivity Method for Computing Total Variation Regularized Fluid Flow
There is a great deal of recent work using optical flow methods for analyzing the dynamics of fluids, and much attention has been paid to developing regularization schemes for variational approaches that are consistent with the physics of fluid flow. In this work we show that using total variation to regularize two different kinds of optical flow functionals leads to very good flow field reconstructions for the kinds of dynamical structures that appear in fluid flow. The first optical flow functional is the classical component-based conservation of intensity, and the second approach is to reconstruct the potential of the flow, rather than the flow components. In the two cases, total variation regularization corresponds to imposing different scientific priors on the solution, which we compute using a variation of the Lagged Diffusivity Fixed Point Iteration. Numerical details are presented, and the results are demonstrated on synthetic data and on a data-driven oceanic flow model.
- Research Organization:
- Nevada Test Site (NTS), Mercury, NV (United States)
- Sponsoring Organization:
- USDOE National Nuclear Security Administration (NNSA)
- DOE Contract Number:
- AC52-06NA25946
- OSTI ID:
- 1136516
- Report Number(s):
- DOE/NV/25946-1553; ISBN 978-0-8218-8737-0
- Journal Information:
- Contemporary Mathematics, Vol. 586; Conference: Eighth International Conference on Scientific Computing and Applications, University of Nevada, Las Vegas, Nevada, April 1-4, 2012; ISSN 0271-4132
- Country of Publication:
- United States
- Language:
- English
Similar Records
Cerebral perfusion computed tomography deconvolution via structure tensor total variation regularization
Recovering fine details from under-resolved electron tomography data using higher order total variation ℓ1 regularization