This window consists in two subregions (the last one is empty),
the one at the top contains nonlinear options and
linear options and the following has degradation
options. At the very bottom there is a Help line that shows
the purpose of the control the mouse is on.
Nonlinear options
Depending on the method we have the following nonlinear options:
- Primal/Primal-dual
- Beta stop:
Beta stop is the regularizing parameter appearing in the
modified TV functional.
- Beta start:
Selects the initial beta in the continuation procedure.
- Beta step:
Selects the factor by which beta is decreased after a
successful subproblem solution in the continuation procedure.
- NWT tol:
Tolerance for the (outer) iteration.
- NWT its:
Maximum number of outer iterations.
- Fixed point
- Beta:
Beta is the regularizing parameter appearing in the
modified TV functional.
- FP tol:
Tolerance for the (outer) iteration.
- FP its:
Maximum number of outer iterations.
- Time marching
- Beta:
Beta is the regularizing parameter appearing in the
modified TV functional.
- Time step:
Initial time step length to ensure stability and sufficient
decrease.
- TM tol:
Tolerance for the (outer) iteration.
- TM its:
Maximum number of outer iterations.
Linear options
The following linear options are available for the Newton-type and
fixed point methods:
- CG rel.tol:
relative tolerance on residual for CG stopping criterion.
- CG its:
maximum number of iterations for CG.
- Preconditioner:
selects the preconditioner for CG. Depending on whether there
is blur or not we have the following options:
No blur:
- No:
no preconditioner.
- ILU:
incomplete Cholesky preconditioner (default).
- DS:
diagonal scaling preconditoner.
Blur:
- No:
no preconditioner.
- CT+DS:
Fast cosine transform preconditioner, combined
with diagonally scaling(default).
- CT:
Fast cosine transform preconditioner.
Degradation options
The popup menu at the left of the degradation options region selects
the type of blur. The currently supported blurs are:
- None
- Gaussian:
G(i,j)=e
(i 2+ j 2)/(4*Theta) /K, for
i=-Width, ..., Width, j=-Height, ..., Height and with K=sum(G(i,j)).
Selecting this will trigger the appearance of
three more windows to select the parameters above
- Width
- Height
- Theta
- Custom:
user supplied PSF, which is loaded by
clicking on the Load button that appears after selecting
this type of blur.
The next popup is used to choose between the two restoration
strategies:
- Tikhonov:
need to supply the alpha parameter in the
corresponding control.
- Constraints:
need to supply an estimate of the variance of
the noise in the corresponding control.
The SNR control allows to specify some amount of noise so that it
matches the entered number. It is tightly related to the Variance
control.
|