Note: This document is for an older version of GRASS GIS that will be discontinued soon. You should upgrade, and read the current manual page.
NAME
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.
KEYWORDS
raster,
algebra,
statistics,
aggregation,
neighbor,
focal statistics,
filter
SYNOPSIS
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]
Flags:
- -a
- Do not align output with the input
- -c
- Use circular neighborhood
- --overwrite
- Allow output files to overwrite existing files
- --help
- Print usage summary
- --verbose
- Verbose module output
- --quiet
- Quiet module output
- --ui
- Force launching GUI dialog
Parameters:
- input=name [required]
- Name of input raster map
- selection=name
- Name of an input raster map to select the cells which should be processed
- output=name[,name,...] [required]
- Name for output raster map
- method=string[,string,...]
- Neighborhood operation
- Options: average, median, mode, minimum, maximum, range, stddev, sum, count, variance, diversity, interspersion, quart1, quart3, perc90, quantile
- Default: average
- size=integer
- Neighborhood size
- Default: 3
- title=phrase
- Title for output raster map
- weight=name
- Text file containing weights
- gauss=float
- Sigma (in cells) for Gaussian filter
- quantile=float[,float,...]
- Quantile to calculate for method=quantile
- Options: 0.0-1.0
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 1
selection 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 1
Without 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:
- average
- The average value within the neighborhood.
In the following example, the result would be:
(7*4 + 6 + 5 + 4*3)/9 = 5.6667
The result is rounded to the nearest integer (in this case 6).
Raw Data Operation New Data
+---+---+---+ +---+---+---+
| 7 | 7 | 5 | | | | |
+---+---+---+ average +---+---+---+
| 4 | 7 | 4 |--------->| | 6 | |
+---+---+---+ +---+---+---+
| 7 | 6 | 4 | | | | |
+---+---+---+ +---+---+---+
- median
- The value found half-way through a list of the
neighborhood's values,
when these are ranged in numerical order.
- mode
- The most frequently occurring value in the neighborhood.
- minimum
- The minimum value within the neighborhood.
- maximum
- The maximum value within the neighborhood.
- range
- The range value within the neighborhood.
- stddev
- The statistical standard deviation of values
within the neighborhood (rounded to the nearest integer).
- sum
- The sum of values within the neighborhood.
- count
- The count of filled (not NULL) cells.
- variance
- The statistical variance of values
within the neighborhood (rounded to the nearest integer).
- diversity
- The number of different values within the neighborhood.
In the above example, the diversity is 4.
- interspersion
- The percentage of cells containing values which differ from the values
assigned to the center cell in the neighborhood, plus 1.
In the above example, the interspersion is:
5/8 * 100 + 1 = 63.5
The result is rounded to the nearest integer (in this case 64).
- quart1, quart3
- The result will be the first or the third quartile (equal of 25th and 75th percentiles).
- perc90
- The result will be the 90th percentile of neighborhood.
- quantile
- Any quantile as specified by "quantile" input parameter.
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.txt
The contents of the weight.txt file:
This 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:
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
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.
- -a
- If specified, r.neighbors will not align the output
raster map layer with that of the input raster map layer.
The r.neighbors program works in the current geographic region.
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, if unspecified,
r.neighbors will align these geographic region settings.
- -c
-
This flag will use a circular neighborhood for the moving analysis window,
centered on the current cell.
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:
- Whether the input map is integer or floating-point.
- Whether the weighted or unweighted version of the aggregate is used.
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:
- Output is always float: average, variance, stddev, quantiles (with
interpolation).
- Output is always integer: diversity, interspersion.
- Output is integer if unweighted, float if weighted: count.
- Output matches input: minimum, maximum, range, mode (subject to
note 2 above), quantiles (without interpolation).
- Output is integer for integer input and unweighted aggregate,
otherwise float: sum.
Set up 10x10 computational region to aid visual inspection of results
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
SOURCE CODE
Available at:
r.neighbors source code
(history)
Latest change: Mon Nov 18 20:15:32 2019 in commit: 1a1d107e4f6e1b846f9841c2c6fabf015c5f720d
Note: This document is for an older version of GRASS GIS that will be discontinued soon. You should upgrade, and read the current manual page.
Main index |
Raster index |
Topics index |
Keywords index |
Graphical index |
Full index
© 2003-2023
GRASS Development Team,
GRASS GIS 7.8.8dev Reference Manual