r.texture(1grass) GRASS GIS User's Manual r.texture(1grass)
NAME
r.texture - Generate images with textural features from a raster map.
KEYWORDS
raster, algebra, statistics, texture
SYNOPSIS
r.texture
r.texture --help
r.texture [-san] input=name output=basename [size=value] [dis-
tance=value] [method=string[,string,...]] [--overwrite] [--help]
[--verbose] [--quiet] [--ui]
Flags:
-s
Separate output for each angle (0, 45, 90, 135)
Angles are counterclockwise from east: 0 is East to West, 45 is
North-East to South-West
-a
Calculate all textural measurements
-n
Allow NULL cells in a moving window
This will also avoid cropping along edges of the current region
--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
output=basename [required]
Name for output basename raster map(s)
size=value
The size of moving window (odd and >= 3)
Default: 3
distance=value
The distance between two samples (>= 1)
The distance must be smaller than the size of the moving window
Default: 1
method=string[,string,...]
Textural measurement method
Options: asm, contrast, corr, var, idm, sa, sv, se, entr, dv, de,
moc1, moc2
DESCRIPTION
r.texture creates raster maps with textural features from a user-speci-
fied raster map layer. The module calculates textural features based on
spatial dependence matrices at 0, 45, 90, and 135 degrees.
In order to take into account the scale of the texture to be measured,
r.texture allows the user to define the size of the moving window and
the distance at which to compare pixel grey values. By default the
module averages the results over the 4 orientations, but the user can
also request output of the texture variables in 4 different orienta-
tions (flag -s). Please note that angles are defined in degrees of east
and they increase counterclockwise, so 0 is East - West, 45 is
North-East - South-West, 90 is North - South, 135 is North-West -
South-East.
The user can either chose one or several texture measures (see below
for their description) using the method parameter, or can request the
creating of maps for all available methods with the -a.
r.texture assumes grey levels ranging from 0 to 255 as input. The in-
put is automatically rescaled to 0 to 255 if the input map range is
outside of this range. In order to reduce noise in the input data
(thus generally reinforcing the textural features), and to speed up
processing, it is recommended that the user recode the data using
equal-probability quantization. Quantization rules for r.recode can be
generated with r.quantile -r using e.g 16 or 32 quantiles (see example
below).
NOTES
Texture is a feature of specific land cover classes in satellite im-
agery. It is particularly useful in situations where spectral differ-
ences between classes are small, but classes are distinguishable by
their organisation on the ground, often opposing natural to human-made
spaces: cultivated fields vs meadows or golf courses, palm tree planta-
tions vs natural rain forest, but texture can also be a natural phe-
nomen: dune fields, different canopies due to different tree species.
The usefulness and use of texture is highly dependent on the resolution
of satellite imagery and on the scale of the human intervention or the
phenomenon that created the texture (also see the discussion of scale
dependency below). The user should observe the phenomenon visually in
order to determine an adequat setting of the size parameter.
The output of r.texture can constitute very useful additional variables
as input for image classification or image segmentation (object recog-
nition). It can be used in supervised classification algorithms such
as i.maxlik or i.smap, or for the identification of objects in i.seg-
ment, and/or for the characterization of these objects and thus, for
example, as one of the raster inputs of the i.segment.stats addon.
In general, several variables constitute texture: differences in grey
level values, coarseness as scale of grey level differences, presence
or lack of directionality and regular patterns. A texture can be char-
acterized by tone (grey level intensity properties) and structure (spa-
tial relationships). Since textures are highly scale dependent, hierar-
chical textures may occur.
r.texture uses the common texture model based on the so-called grey
level co-occurrence matrix as described by Haralick et al (1973). This
matrix is a two-dimensional histogram of grey levels for a pair of pix-
els which are separated by a fixed spatial relationship. The matrix ap-
proximates the joint probability distribution of a pair of pixels. Sev-
eral texture measures are directly computed from the grey level co-oc-
currence matrix.
The following part offers brief explanations of the Haralick et al tex-
ture measures (after Jensen 1996).
First-order statistics in the spatial domain
• Sum Average (SA)
• Entropy (ENT): This measure analyses the randomness. It is high
when the values of the moving window have similar values. It is
low when the values are close to either 0 or 1 (i.e. when the
pixels in the local window are uniform).
• Difference Entropy (DE)
• Sum Entropy (SE)
• Variance (VAR): A measure of gray tone variance within the mov-
ing window (second-order moment about the mean)
• Difference Variance (DV)
• Sum Variance (SV)
Note that measures "mean", "kurtosis", "range", "skewness", and "stan-
dard deviation" are available in r.neighbors.
Second-order statistics in the spatial domain
The second-order statistics texture model is based on the so-called
grey level co-occurrence matrices (GLCM; after Haralick 1979).
• Angular Second Moment (ASM, also called Uniformity): This is a
measure of local homogeneity and the opposite of Entropy. High
values of ASM occur when the pixels in the moving window are
very similar.
Note: The square root of the ASM is sometimes used as a texture
measure, and is called Energy.
• Inverse Difference Moment (IDM, also called Homogeneity): This
measure relates inversely to the contrast measure. It is a di-
rect measure of the local homogeneity of a digital image. Low
values are associated with low homogeneity and vice versa.
• Contrast (CON): This measure analyses the image contrast (lo-
cally gray-level variations) as the linear dependency of grey
levels of neighboring pixels (similarity). Typically high, when
the scale of local texture is larger than the distance.
• Correlation (COR): This measure analyses the linear dependency
of grey levels of neighboring pixels. Typically high, when the
scale of local texture is larger than the distance.
• Information Measures of Correlation (MOC)
• Maximal Correlation Coefficient (MCC)
The computational region should be set to the input map with g.region
raster=<input map>, or aligned to the input map with g.region
align=<input map> if only a subregion should be analyzed.
Note that the output of r.texture will always be smaller than the cur-
rent region as only cells for which there are no null cells and for
which all cells of the moving window are within the current region will
contain a value. The output will thus appear cropped at the margins.
Importantly, the input raster map cannot have more than 255 categories.
EXAMPLE
Calculation of Angular Second Moment of B/W orthophoto (North Carolina
data set):
g.region raster=ortho_2001_t792_1m -p
# set grey level color table 0% black 100% white
r.colors ortho_2001_t792_1m color=grey
# extract grey levels
r.mapcalc "ortho_2001_t792_1m.greylevel = ortho_2001_t792_1m"
# texture analysis
r.texture ortho_2001_t792_1m.greylevel prefix=ortho_texture method=asm -s
# display
g.region n=221461 s=221094 w=638279 e=638694
d.shade color=ortho_texture_ASM_0 shade=ortho_2001_t792_1m
This calculates four maps (requested texture at four orientations): or-
tho_texture_ASM_0, ortho_texture_ASM_45, ortho_texture_ASM_90, or-
tho_texture_ASM_135. Reducing the number of gray levels (equal-proba-
bility quantizing):
g.region -p raster=ortho_2001_t792_1m
# enter as one line or with \
r.quantile input=ortho_2001_t792_1m quantiles=16 -r | r.recode \
input=ortho_2001_t792_1m output=ortho_2001_t792_1m_q16 rules=-
The recoded raster map can then be used as input for r.texture as be-
fore.
Second example: analysis of IDM (homogeneity) on a simple raster with
North-South line pattern.
# import raster
r.in.ascii in=- output=lines << EOF
north: 9
south: 0
east: 9
west: 0
rows: 9
cols: 9
0 0 0 1 0 0 0 1 0
0 0 0 1 0 0 0 1 0
0 0 0 1 0 0 0 1 0
0 0 0 1 0 0 0 1 0
0 0 0 1 0 0 0 1 0
0 0 0 1 0 0 0 1 0
0 0 0 1 0 0 0 1 0
0 0 0 1 0 0 0 1 0
0 0 0 1 0 0 0 1 0
EOF
# adjust region to raster
g.region raster=lines
# calculate IDM (homogeneity) in all directions
r.texture -s lines method=idm output=text_lines
The following image shows the original map, the result in East-West di-
rection and the result in North-South direction, showing how texture
can depend on direction, with texture perfectly homogeneous (value=1)
in the North-South direction, but quite heterogeneous in East-West di-
rection, except for those areas where there are three columns of equal
values (as size=3). The overlaid grid highlights that the texture mea-
sures output maps are cropped at the margins.
IDM textures according to direction
KNOWN ISSUES
The program can run incredibly slow for large raster maps and large
moving windows (size option).
REFERENCES
The algorithm was implemented after Haralick et al., 1973 and 1979.
The original code was taken by permission from pgmtexture, part of PBM-
PLUS (Copyright 1991, Jef Poskanser and Texas Agricultural Experiment
Station, employer for hire of James Darrell McCauley). Manual page of
pgmtexture. Over the years, the source code of r.texture was further
improved.
• Haralick, R.M., K. Shanmugam, and I. Dinstein (1973). Textural
features for image classification. IEEE Transactions on Sys-
tems, Man, and Cybernetics, SMC-3(6):610-621.
• Bouman, C. A., Shapiro, M. (1994). A Multiscale Random Field
Model for Bayesian Image Segmentation, IEEE Trans. on Image
Processing, vol. 3, no. 2.
• Jensen, J.R. (1996). Introductory digital image processing.
Prentice Hall. ISBN 0-13-205840-5
• Haralick, R. (May 1979). Statistical and structural approaches
to texture, Proceedings of the IEEE, vol. 67, No.5, pp. 786-804
• Hall-Beyer, M. (2007). The GLCM Tutorial Home Page (Grey-Level
Co-occurrence Matrix texture measurements). University of Cal-
gary, Canada
SEE ALSO
i.maxlik, i.gensig, i.smap, i.gensigset, i.segment.stats, i.pca,
r.neighbors, r.rescale
AUTHORS
G. Antoniol - RCOST (Research Centre on Software Technology - Viale
Traiano - 82100 Benevento)
C. Basco - RCOST (Research Centre on Software Technology - Viale Tra-
iano - 82100 Benevento)
M. Ceccarelli - Facolta di Scienze, Universita del Sannio, Benevento
Markus Metz (correction and optimization of the initial version)
Moritz Lennert (documentation)
SOURCE CODE
Available at: r.texture source code (history)
Accessed: unknown
Main index | Raster index | Topics index | Keywords index | Graphical
index | Full index
© 2003-2022 GRASS Development Team, GRASS GIS 7.8.7 Reference Manual
GRASS 7.8.7 r.texture(1grass)
Generated by dwww version 1.14 on Fri Jan 24 09:46:16 CET 2025.