Class VariationalRefinement


public class VariationalRefinement
extends DenseOpticalFlow
Variational optical flow refinement This class implements variational refinement of the input flow field, i.e. it uses input flow to initialize the minimization of the following functional: \(E(U) = \int_{\Omega} \delta \Psi(E_I) + \gamma \Psi(E_G) + \alpha \Psi(E_S) \), where \(E_I,E_G,E_S\) are color constancy, gradient constancy and smoothness terms respectively. \(\Psi(s^2)=\sqrt{s^2+\epsilon^2}\) is a robust penalizer to limit the influence of outliers. A complete formulation and a description of the minimization procedure can be found in CITE: Brox2004