OOF2: The Manual
TotalVariation (TotalVariation) |
||
|---|---|---|
|
6.5.2. Subclasses |  |
Name
TotalVariation (TotalVariation) — Denoise using total variation regularization, suitable for piecewise constant images.
Synopsis
TotalVariation(weight, eps, max_iterations)
Details
-
Base class:
DenoiseMethod -
Parameters:
weight- Larger values remove more noise, at the expense of image fidelity. Type: A real number.
eps- Stop iterating when the estimated error is below this value. Type: A positive real number.
max_iterations- Maximum number of iterations. Type: Integer.
Description
TotalVariation can be used as the
method parameter to the OOF.Image.Modify.Denoise command. It calls
skimage.restoration.denoise_tv_chambolle
from the scikit-image library. It finds an image similar to the
original image but with less variation, by finding values
that minimize the function
where the sum is over pixels , is the value of pixel
, and is the optimized value of
the image at . is the
weight parameter. The weight
governs the relative importance of smoothness and fidelity (to the
original image) in the denoised image. A large
weight will produce less noise and also less
fidelity.
The parameter eps governs how hard the algorithm
tries to find an optimal solution. The algorithm stops when the
estimated relative error is less than eps, or
when it has used more than max_iterations
iterations. Smaller values give better solutions and take longer to
run.
Figure 6.123. Denoising with TotalVariation
The top left image is a 200×200 pixel subsection of an
Escher print. The top right is the same section, with added
noise. The other three images are the result of denoising
using TotalVariation with the given
weight values. The e
value in each image is the norm of the difference between it
and the original image. (The norm may or may not be a useful
measure of the effectiveness of the procedure.)
|
|
 |
|
Other Mesh (ThreeVectorOtherMeshInit)
|
 |
TransitionPoints (TransitionPoints) |