SUSAN Noise Reduction |
SUSAN noise reduction uses nonlinear filtering to reduce noise in
an image (2D or 3D) whilst preserving the underlying structure. It
does this by only averaging a voxel with local voxels which have
similar intensity. For further detail, see here. To
reference SUSAN, quote
S.M. Smith and J.M. Brady. SUSAN - a new
approach to low level image processing. Int. Journal of Computer
Vision, 23(1):45-78, May 1997.
Dimensionality: this should be set by default to the dimensionality (2D or 3D) of the image currently selected when the SUSAN GUI was opened. If the image to be processed is 3D, slices can be processed separately (i.e., set dimensionality to 2D) or the processing can be volumetric (3D).
Brightness threshold: this allows SUSAN to discriminate between noise and the underlying image. Ideally, the value should be set greater than the noise level and less than the contrast of the underlying image. Edges of contrast smaller than this threshold will tend to be blurred by SUSAN whereas those of greater contrast will not be.
Mask SD: this determines the spatial extent of the smoothing. The mask is basically Gaussian with standard deviation (in image units - e.g. mm) set by the user. However, for a small, fast, flat response 3x3 or 3x3x3 voxel mask, set SD to 0.
Use median when no neighbourhood is found: by default, when the local neighbourhood of similar brightness voxels is empty, a local median filter is used. This allows the correction of impulse ("salt-and-pepper") noise. This feature can be turned off if desired. In this case, when no neighbourhood is found, the original intensity of the voxel of interest remains unchanged.
Separate images to find USAN from: if the local area to be smoothed over is to be found from an image (or images) other than the main input image, this can be input here. The USAN images should be pasted from the page manager.