The program will be run non-interactively if the user specifies program arguments (see OPTIONS) on the command line. Alternately, the user can simply type r.neighbors on the command line, without program arguments. In this case, the user will be prompted for flag settings and parameter values.
The user can optionally specify a selection map, to compute new values only where the raster cells of the selection map are not NULL. In case of a NULL cells, the values from the input map are copied into the output map. This may useful to smooth only parts of an elevation map (pits, peaks, ...).
Example how to use a selection map with method=average:
input map:
1 1 1 1 1 1 1 1 1 1 1 1 10 1 1 1 1 1 1 1 1 1 1 1 1selection map, NULL values are marked as *:
* * * * * * * 1 * * * 1 1 1 * * * 1 * * * * * * *The output map:
1 1 1 1 1 1 1 2 1 1 1 2 2 2 1 1 1 2 1 1 1 1 1 1 1Without using the selection map, the output map would look like this:
1 1 1 1 1 1 2 2 2 1 1 2 2 2 1 1 2 2 2 1 1 1 1 1 1
Optionally, the user can also specify the TITLE to be assigned to the raster map layer output, elect to not align the resolution of the output with that of the input (the -a option), and run r.neighbors with a custom matrix weights with the weight option. These options are described further below.
Neighborhood Operation Methods: The neighborhood operators determine what new value a center cell in a neighborhood will have after examining values inside its neighboring cells. Each cell in a raster map layer becomes the center cell of a neighborhood as the neighborhood window moves from cell to cell throughout the map layer. r.neighbors can perform the following operations:
Raw Data Operation New Data ---------------- ---------------- | 7 | 7 | 5 | | | | | |----|----|----| average |----|----|----| | 4 | 7 | 4 |--------->| | 6 | | |----|----|----| |----|----|----| | 7 | 6 | 4 | | | | | |----|----|----| |----|----|----|
Neighborhood Size:
The neighborhood size specifies which cells surrounding any given
cell fall into the neighborhood for that cell.
The size must be an odd integer.
For example,
_ _ _ |_|_|_| 3 x 3 neighborhood ---> |_|_|_| |_|_|_|
Matrix weights: A custom matrix can be used if none of the neighborhood operation methods are desirable by using the weight. This option must be used in conjunction with the size option to specify the matrix size. The weights desired are to be entered into a text file. For example, to calculate the focal mean with a matrix size of 3,
r.neigbors in=input.map out=output.map size=3 weight=weights.txtThe contents of the weight.txt file:
3 3 3 1 4 8 9 5 3This corresponds to the following 3x3 matrix:
------- |3|3|3| ------- |1|4|8| ------- |9|5|3| -------The way that weights are used depends upon the specific aggregate (method) being used. However, most of the aggregates have the property that multiplying all of the weights by the same factor won't change the final result (an exception is method=count). Also, most (if not all) of them have the properties that an integer weight of N is equivalent to N occurrences of the cell value, and having all weights equal to one produces the same result as when weights are not used. When weights are used, the calculation for method=average is:
sum(w[i]*x[i]) / sum(w[i])In the case where all weights are zero, this will end up with both the numerator and denominator to zero, which produces a NULL result.
The exact masks for the first few neighborhood sizes are as follows:
3x3 . X . 5x5 . . X . . 7x7 . . . X . . . X O X . X X X . . X X X X X . . X . X X O X X . X X X X X . . X X X . X X X O X X X . . X . . . X X X X X . . X X X X X . . . . X . . . 9x9 . . . . X . . . . 11x11 . . . . . X . . . . . . . X X X X X . . . . X X X X X X X . . . X X X X X X X . . X X X X X X X X X . . X X X X X X X . . X X X X X X X X X . X X X X O X X X X . X X X X X X X X X . . X X X X X X X . X X X X X O X X X X X . X X X X X X X . . X X X X X X X X X . . . X X X X X . . . X X X X X X X X X . . . . . X . . . . . X X X X X X X X X . . . X X X X X X X . . . . . . . X . . . . .
r.neighbors doesn't propagate NULLs, but computes the aggregate over the non-NULL cells in the neighborhood.
The -c flag and the weights parameter are mutually exclusive. Any use of the two together will produce an error. Differently-shaped neighborhood analysis windows may be achieved by using the weight= parameter to specify a weights file where all values are equal. The user can also vary the weights at the edge of the neighborhood according to the proportion of the cell that lies inside the neighborhood circle, effectively anti-aliasing the analysis mask.
For aggregates where a weighted calculation isn't meaningful (specifically: minimum, maximum, diversity and interspersion), the weights are used to create a binary mask, where zero causes the cell to be ignored and any non-zero value causes the cell to be used.
r.neighbors copies the GRASS color files associated with the input raster map layer for those output map layers that are based on the neighborhood average, median, mode, minimum, and maximum. Because standard deviation, variance, diversity, and interspersion are indices, rather than direct correspondents to input values, no color files are copied for these map layers. (The user should note that although the color file is copied for average neighborhood function output, whether or not the color file makes sense for the output will be dependent on the input data values.)
Last changed: $Date$