next up previous contents
Next: Muti-grid approach and optimization Up: Description of the algorithm Previous: Cost function   Contents

Regularization

The following regularization terms were used in our numerical experiments:

$\displaystyle R_0(u,v)$ $\displaystyle =$ $\displaystyle \Vert u\Vert^2+\Vert v\Vert^2,$ (4.4)
$\displaystyle R_1(u,v)$ $\displaystyle =$ $\displaystyle \Vert\nabla u\Vert^2+\Vert\nabla v\Vert^2=\Vert\partial_x u\Vert^2+\Vert\partial_y u\Vert^2 + \Vert\partial_x v\Vert^2+\Vert\partial_y v\Vert^2,$ (4.5)
$\displaystyle R_{div}(u,v)$ $\displaystyle =$ $\displaystyle \Vert div(u,v)\Vert^2=\Vert\partial_x u+\partial_y v\Vert^2,$ (4.6)
$\displaystyle R_{curl}(u,v)$ $\displaystyle =$ $\displaystyle \Vert curl(u,v)\Vert^2=\Vert\partial_y u-\partial_x v\Vert^2,$ (4.7)
$\displaystyle R_{div/curl}(u,v)$ $\displaystyle =$ $\displaystyle \Vert div(u,v)\Vert^2+\Vert curl(u,v)\Vert^2 = \Vert\partial_x u+\partial_y v\Vert^2+\Vert\partial_y u-\partial_x v\Vert^2,$ (4.8)
$\displaystyle R_{\nabla div}(u,v)$ $\displaystyle =$ $\displaystyle \Vert\nabla div(u,v)\Vert^2=\Vert\partial^2_{xx} u+\partial^2_{xy} v\Vert^2+\Vert\partial^2_{xy} u+\partial^2_{yy} v\Vert^2,$ (4.9)
$\displaystyle R_{\nabla div/\nabla curl}(u,v)$ $\displaystyle =$ $\displaystyle \Vert\nabla div(u,v)\Vert^2+\Vert\nabla curl(u,v)\Vert^2$ (4.10)
  $\displaystyle =$ $\displaystyle \Vert\partial^2_{xx} u+\partial^2_{xy} v\Vert^2+\Vert\partial^2_{...
...xy} u-\partial^2_{xx} v\Vert^2+\Vert\partial^2_{yy} u-\partial^2_{xy} v\Vert^2.$  

In all the cases, we can write $ R(u,v)=\Vert S(u,v)\Vert^2$ , where $ S$ is a linear operator. Some scalar coefficients have also been considered in order to weight the different terms of a given regularization.


next up previous contents
Next: Muti-grid approach and optimization Up: Description of the algorithm Previous: Cost function   Contents
Back to home page