Filters (Spatial): Nonlinear Noise Reduction
From MIPAV
Background
This algorithm provides a port of the SUSAN (Smallest Univalue Segment Assimilating Nucleus) lowlevel image processing program.
For information about SUSAN, see http://www.fmrib.ox.ac.uk/~steve.
SUSAN noise reduction uses nonlinear filtering to reduce noise in an image while preserving both the underlying structure and the edges and corners. It does this by averaging a voxel only with local voxels that have similar intensity.
Table 1 explains the possible responses that the SUSAN filter provides if it cannot determine the neighborhood and based on whether you selected the option of using the median.
Use median

SUSAN area

The SUSAN filter . . .

Selected

Zero

Uses the median of the pixel's 8 nearest neighbors in 2D images or 26 nearest neighbors in 3D images to estimate the pixel's correct value.

Selected

Nonzero

Uses sums taken over the local neighborhood, not including the center pixel itself, to estimate the pixel's correct value. This allows good reduction of impulse noise.

Not selected

Nonzero

Uses sums taken over the local neighborhood including the center pixel to estimate the pixel's correct value.

Not selected

Zero

Leaves the pixel unchanged.

Image types
You can apply this algorithm to blackandwhite 2D, 2.5D, and 3D images.
Special notes
None.
Applying the algorithm
To run this algorithm, complete the following steps:
 Open an image.
 Select Algorithms > Filter (spatial) > Nonlinear noise reduction. The Nonlinear Noise Reduction dialog box (Figure 1) opens.
 Complete the information in the dialog box.
 Click OK. The algorithm begins to run.
 A popup window appears with the status. The following message appears
 "Performing the nonlinear noise reduction." When the algorithm finishes running, the popup window closes.
 Depending on whether you selected New image or Replace image, the results appear in a new window or replace the image to which the algorithm was applied.
Brightness threshold

Blurs edges of contrast smaller than this threshold. Reducing the brightness threshold gives less smoothing. The default value is 0.1 * (image maximum  image minimum).


Mask Gaussian SD (0 for flat response) mm

Determines the spatial extent of the smoothing (the default value is equal to the X resolution). For a small, fast, flat response with a 3 x 3 or 3 x 3 x 3 voxel mask, set the mask SD to 0.
 
Use median when neighborhood not found

By default, this check box is selected.
 
Process each slice independently

Smooths each slice of the dataset independently (applies only to 3D images).
 
New image

Shows the results of the algorithm in a new image window (default choice).
 
Replace image

Replaces the current active image with the newly calculated image.
 
OK

Applies the algorithm according to the specifications in this dialog box.
 
Cancel

Disregards any changes that you made in the dialog box and closes the dialog box.
 
Help

Displays online help for this dialog box.

See also:
 Filters (Spatial): Adaptive Noise Reduction
 Filters (Frequency)
 Filters (Spatial): Adaptive Path Smooth
 Filters (Spatial) Anisotropic Diffusion
 Filters (Spatial): CoherenceEnhancing Diffusion
 Filters (Spatial): Gaussian Blur
 Filters (Spatial): Gradient Magnitude
 Filters (Spatial): Haralick Texture
 Filters (Spatial) Laplacian
 Filters (Spatial): Local Normalization
 Filters (Spatial): Mean
 Filters (Spatial): Median
 Filters (Spatial): Mode
 Filters (Spatial): Nonmaximum Suppression
 Filters (Spatial): Regularized Isotropic (Nonlinear) Diffusion
 Filters (Spatial): Slice Averaging
 Filters (Spatial): Unsharp Mask