Set the pixel function of a VRT stack object

Usage

vrt_set_py_pixelfun(x, pixfun, band_idx)

# S3 method for class 'vrt_block'
vrt_set_py_pixelfun(x, pixfun = vrtility::median_numpy(), band_idx = NULL)

# S3 method for class 'vrt_collection'
vrt_set_py_pixelfun(x, pixfun = vrtility::median_numpy(), band_idx = NULL)

median_numpy()

mean_numpy()

geomean_numpy()

quantile_numpy(q, use_fastnanquantile = TRUE)

mean_db_numpy()

Arguments

x: A vrt_stack object
pixfun: A function that returns the Python code for the pixel function
band_idx: The indices of the bands to set the pixel function for. If NULL, the pixel function is set for all bands.
q: Probability of the quantile to compute. Values must be between 0 and 1 inclusive.
use_fastnanquantile: Logical indicating whether to use the fastnanquantile library for calculating the quantile. This is generally faster than the standard numpy implementation, especially for large arrays. However, it requires the fastnanquantile library to be installed in the Python environment (This is handled automatically). Default is TRUE.

Value

character of the python function

Details

median_numpy is a pixel function that calculates the median of the input arrays, it is injected into the VRT file as a Python function. mean_numpy works in the same way but calculates the mean.`

geomean_numpy is a pixel function that calculates the geometric mean of the input arrays. Use cases of this are at present unclear to me. If you have thoughts or references please let me know.

quantile_numpy is a pixel function that calculates the quantile of the input arrays for a given probability. This could be useful where the median fails to filter cloudy pixels effectively. The defauly numpy nanquantile function is very slow and does not support masked arrays. Therefore, it is typically much faster to use the fastnanquantile library which uses numba to increase performance. However, this library has some limitations around multi-threading in certain contexts. Therefore, when using this pixel function with vrt_compute(), it is strongly advised to use the warp engine and to avoid using mirai daemons (i.e. use sequential processing).

mean_db_numpy is a pixel function that calculates the mean of the input arrays and then converts to decibels. This is useful for calculating the mean of radar raw/linear backscatter values, for example.