Simple raster data processing on the command line is possible using Rasterio’s rio-calc command. It uses the snuggs Numpy S-expression engine. The snuggs README explains how expressions are written and evaluated in general. This document explains Rasterio-specific details of rio-calc and offers some examples.
Rio-calc expressions look like
(func|operator arg [*args])
func may be the name of any function in the module
one of the rio-calc builtins:
operator may be any of the standard Python arithmetic or logical operators.
The arguments may themselves be expressions.
Copying a file
Here’s a trivial example of copying a dataset. The expression
evaluates to all bands of the first input dataset, an array with shape
(3, 718, 791) in this case.
Note: rio-calc’s indexes start at
$ rio calc "(read 1)" tests/data/RGB.byte.tif out.tif
Reversing the band order of a file
(read i j) evaluates to the j-th band of the i-th input
asarray function collects bands read in reverse order into
an array with shape
(3, 718, 791) for output.
$ rio calc "(asarray (read 1 3) (read 1 2) (read 1 1))" tests/data/RGB.byte.tif out.tif
Stacking bands of multiple files
Bands can be read from multiple input files. This example is another (slower) way to copy a file.
$ rio calc "(asarray (read 1 1) (read 2 2) (read 3 3))" \ > tests/data/RGB.byte.tif tests/data/RGB.byte.tif tests/data/RGB.byte.tif \ > out.tif
Datasets can be referenced in expressions by name and single bands picked out
$ rio calc "(asarray (take a 3) (take a 2) (take a 1))" \ > --name "a=tests/data/RGB.byte.tif" out.tif
The third example, re-done using names, is:
$ rio calc "(asarray (take a 1) (take b 2) (take b 3))" \ > --name "a=tests/data/RGB.byte.tif" --name "b=tests/data/RGB.byte.tif" \ > --name "c=tests/data/RGB.byte.tif" out.tif
Read and take
take overlap a bit in the previous examples but
are rather different. The former involves I/O and the latter does not. You may
take from any array, as in this example.
$ rio calc "(take (read 1) 1)" tests/data/RGB.byte.tif out.tif
Arithmetic operations can be performed as with Numpy. Here is an example of scaling all three bands of a dataset by the same factors.
$ rio calc "(+ 2 (* 0.95 (read 1)))" tests/data/RGB.byte.tif out.tif
Here is a more complicated example of scaling bands by different factors.
$ rio calc "(asarray (+ 2 (* 0.95 (read 1 1))) (+ 3 (* 0.9 (read 1 2))) (+ 4 (* 0.85 (read 1 3))))" tests/data/RGB.byte.tif out.tif
Logical operations can be used in conjunction with arithemtic operations. In this example, the output values are 255 wherever the input values are greater than or equal to 40.
$ rio calc "(* (>= (read 1) 40) 255)" tests/data/RGB.byte.tif out.tif