next up previous contents index
Next: cccopy Up: Reference Previous: fsym2   Contents   Index

wiener

$ \bigcirc$Name


wiener Wiener Filter (least squares with H1 regularization)




$ \bigcirc$Command Synopsis


wiener [-w lambda] [-K kernel] [-R rad] [-g g] in out



-w lambda : weight on fidelity term (default: 1.0)

-K kernel : specify blur kernel in Fourier domain

-R rad : ... or radial kernel in Fourier domain

-g g : ... or gaussian standart deviation

in : input Fimage

out : output Fimage




$ \bigcirc$Function Summary


Fimage wiener (in , out , kernel , rad , g , lambda )

Fimage in , kernel , out ;

Fsignal rad ;

float *lambda , *g ;




$ \bigcirc$Description


This module implements the classical Wiener filter (in its simplest version), that aims to restore an image in by an image out that minimizes the energy

E(out) = $\displaystyle \int$|$\displaystyle \nabla$(out)|2 + w$\displaystyle \int$(K$\displaystyle \star$out - in)2.

The solution is obtained in Fourier domain with

$\displaystyle \widehat{{out}}$($\displaystyle \xi$) = $\displaystyle \widehat{{in}}$($\displaystyle \xi$) . $\displaystyle {\frac{{w K(\xi)}}{{w K^2(\xi) + \pi^2\vert\xi^2\vert}}}$,

where $ \xi$ is normalized into [- $ {\frac{1}{2}}$,$ {\frac{1}{2}}$]2. The blur kernel K can be




$ \bigcirc$See Also


fft2d.




$ \bigcirc$Version 1.1


Last Modification date : Fri Jan 25 19:32:16 2002


$ \bigcirc$Author


Lionel Moisan






next up previous contents index
Next: cccopy Up: Reference Previous: fsym2   Contents   Index
mw 2004-05-05