r.neighbors(1grass) | GRASS GIS User's Manual | r.neighbors(1grass) |
r.neighbors - Makes each cell category value a function of the category values assigned to the cells around it, and stores new cell values in an output raster map layer.
raster, algebra, statistics, aggregation, neighbor, focal statistics, filter
r.neighbors
r.neighbors --help
r.neighbors [-ac] input=name
[selection=name] output=name[,name,...]
[method=string[,string,...]]
[size=integer] [title=phrase]
[weight=name] [gauss=float]
[quantile=float[,float,...]] [--overwrite]
[--help] [--verbose] [--quiet] [--ui]
r.neighbors looks at each cell in a raster input file, and examines the values assigned to the cells in some user-defined "neighborhood" around it. It outputs a new raster map layer in which each cell is assigned a value that is some (user-specified) function of the values in that cell’s neighborhood. For example, each cell in the output layer might be assigned a value equal to the average of the values appearing in its 3 x 3 cell "neighborhood" in the input layer. Note that the centre cell is also included in the calculation.
The user must specify the names of the raster map layers to be used for input and output, the method used to analyze neighborhood values (i.e., the neighborhood function or operation to be performed), and the size of the neighborhood.
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 and represent the length of one
of moving window edges in cells. For example, a size value of 3 will result
in
_ _ _
|_|_|_|
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| +-+-+-+To calculate an annulus shaped neighborhood the contents of weight.txt file may be e.g. for size=5:
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:
0 1 1 1 0
1 0 0 0 1
1 0 0 0 1
1 0 0 0 1
0 1 1 1 0
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.
sum(w[i]*x[i]) / sum(w[i])
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 . . . . .
The r.neighbors program works in the current geographic region with the current mask, if any. It is recommended, but not required, that the resolution of the geographic region be the same as that of the raster map layer. By default, r.neighbors will align these geographic region settings. However, the user can select to keep original input and output resolutions which are not aligned by specifying this (e.g., using the -a option).
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.)
The following logic has been implemented: For any aggregate, there are two factors affecting the output type:
These combine to create four possibilities:
input type/weight | integer | float | ||||||
no | yes | no | yes | |||||
average | float | float | float | float | ||||
median | [1] | [1] | float | float | ||||
mode | integer | integer | [2] | [2] | ||||
minimum | integer | integer | float | float | ||||
maximum | integer | integer | float | float | ||||
range | integer | integer | float | float | ||||
stddev | float | float | float | float | ||||
sum | integer | float | float | float | ||||
count | integer | float | integer | float | ||||
variance | float | float | float | float | ||||
diversity | integer | integer | integer | integer | ||||
interspersion | integer | integer | integer | integer | ||||
quart1 | [1] | [1] | float | float | ||||
quart3 | [1] | [1] | float | float | ||||
perc90 | [1] | [1] | float | float | ||||
quantile | [1] | [1] | float | float |
[1] For integer input, quantiles may produce float results from
interpolating between adjacent values.
[2] Calculating the mode of floating-point data is essentially
meaningless.
With the current aggregates, there are 5 cases:
Set up 10x10 computational region to aid visual inspection of
results
g.region rows=10 cols=10
Fill 50% of computational region with randomly located cells.
"distance=0" will allow filling adjacent cells.
r.random.cells output=random_cells distance=0 ncells=50
Count non-empty (not NULL) cells in 3x3 neighborhood
r.neighbors input=random_cells output=counts method=count
Optionally - exclude centre cell from the count (= only look
around)
r.mapcalc "cound_around = if( isnull(random_cells), counts, counts - 1)"
g.region
r.clump
r.mapcalc
r.mfilter
r.statistics
r.support
Original version: Michael Shapiro, U.S.Army Construction
Engineering Research Laboratory
Updates for GRASS GIS 7 by Glynn Clements and others
Available at: r.neighbors source code (history)
Main index | Raster index | Topics index | Keywords index | Graphical index | Full index
© 2003-2020 GRASS Development Team, GRASS GIS 7.8.5 Reference Manual
GRASS 7.8.5 |