The Gaussian filter option is the most widely used filter option.
The Gaussian filter option in the Construct Filter Feature dialog box (Insert | Feature | Constructed | Filter) enables you to smooth the data by applying a Linear or Polar Gaussian lowpass filter to it according to the ISO 11562 standard. The Cutoff wavelength, Cutoff frequency, or Filter width value controls the amount of smoothing.
For the polar Gaussian filter, the data should be a complete circle scan, with radial deviations. A partial circle scan will not work correctly with this filter. The appropriate filter for a partial circle scan is the cylindrical filter.
For the linear Gaussian filter, the data should nominally lie in a plane with the deviations perpendicular to the plane. For this filter, the cutoff wavelength is in units of length. The following paragraphs describe how a linear Gauss filter functions:
It estimates the point spacing delta as the average distance between the X,Y coordinates of the 3D points. These should be evenly spaced and coplanar. It will smooth the Z coordinates.
If the smoothing parameter is
m
= filter width
(i.e., the smoothed value is at the center of 2m+1
points used in a weighted average, starting at point m),
then the cutoff wavelength lambda is
computed as:
lambda
= m * delta / const
(where const is a numerical constant.)
If the input parameter is lambda = cutoff wavelength, then the filter width m is computed as:
m = lambda * const / delta (actually, the next higher integer value).
So, the cutoff wavelength has the same units as the point spacing, but the filter width is a pure number.