Introduction

After loading data, you can apply a data filter to smooth or sharpen data. Data filtering is optional, but applying one can remove drastic changes in neighboring response values or clean out false data. Applying a data filter can also highlight or exaggerate the differences between adjacent response values.

For example, suppose you have measurement values from an instrument that has some degree of noise associated with its measurements:

measure value = real value +/- instrument noise

To assist in analysis, you could smooth out the noise by applying a smoothing filter, like the Blend filter. Another example would be a data set representing a CT scan that you want smoothed or enhanced.

When you apply a data filter, the software adjusts the value for each response value in the data by performing a mathematical operation. In general, the operation replaces the response value being operated on by multiplying and averaging its value with the values of adjacent response values. Missing response values are ignored.

When you specify a filter, the software displays three pads of buttons that represent a 3x3x3 matrix. Each button (element) represents a response value location, with the center element representing the response value being operated on. For example, Laplacian Filter Matrix of Preset Values shows the matrix of preset values for the Laplacian filter (provided with the software), which sharpens data.

Laplacian Filter Matrix of Preset Values

[IMAGE]

When you apply the Laplacian filter, the software does the following for each response value in the data:

The response value being operated on is multiplied by 7, which weights (increases) it.
Adjacent (surrounding) response values are multiplied by -1, which pulls down (decreases) their values.
The resulting values are then averaged, and the average replaces the value for the response value being operated on.

Chapter Contents
Previous
Next
Top of Page