Basic Description: Linear is a linear filter operation that can be used to run low-pass filters (average, disk, or Gaussian) to smooth the data, and high pass filters (Laplacian, log, and unsharp) to accent edges and transitions.
Technical Details: A linear filter is an operation where each pixel is transformed by a linear combination of pixels in the surrounding neighborhood. The neighborhood is specified by the user. The linear filter is implemented as a convolution on the tile.
Instructions: After choosing a selection method and making your selection on the survey, select a filter type from the drop-down list and follow the instruction for that filter.
 average
When the average filter is selected, simply choose a kernel size and units (samples or meters) using the drop-down menus.
|
 disk
The disk option averages values within a circular region with the specified radius (in pixels).
|
 gaussian
The gaussian filter is a weighted average of the pixels in the defined neighborhood (i.e. the filter size). The weights of the neighboring pixels decrease with distance from the center and are based on a Gaussian curve with the specified Sigma value. Small values of sigma create a faster drop in weight with increasing distance. Typically the filter size should be chosen as 2*Sigma + 1. For example, if Sigma = 2, the filter size should be 5x5 samples. If Sigma < 1, chose the Filter Size as 3x3.
|
 laplacian
The laplacian filter is an edge sharpening filter and is sometimes called a top-hat filter because of its shape. Alpha controls the amount of sharpening with values close to one resulting in extreme, and often undesired, sharpening.
|
 log
The LoG, or Laplacian of Gaussian filter, is another edge sharpening filter. It is identical to the Laplacian filter but the additional capacity for smoothing the tile first with a Gaussian low-pass filter. This is useful for emphasizing edges at different scales.
|
 unsharp
Unsharp is a sharping filter commonly used in image processing. Large alpha values result in more sharpening.
|
