astropy.nddata.extract_array(array_large, shape, position, mode='partial', fill_value=nan, return_position=False)[source]#

Extract a smaller array of the given shape and position from a larger array.


The array from which to extract the small array.

shapeint or tuple thereof

The shape of the extracted array (for 1D arrays, this can be an int). See the mode keyword for additional details.

positionnumber or tuple thereof

The position of the small array’s center with respect to the large array. The pixel coordinates should be in the same order as the array shape. Integer positions are at the pixel centers (for 1D arrays, this can be a number).

mode{‘partial’, ‘trim’, ‘strict’}, optional

The mode used for extracting the small array. For the 'partial' and 'trim' modes, a partial overlap of the small array and the large array is sufficient. For the 'strict' mode, the small array has to be fully contained within the large array, otherwise an PartialOverlapError is raised. In all modes, non-overlapping arrays will raise a NoOverlapError. In 'partial' mode, positions in the small array that do not overlap with the large array will be filled with fill_value. In 'trim' mode only the overlapping elements are returned, thus the resulting small array may be smaller than the requested shape.

fill_valuenumber, optional

If mode='partial', the value to fill pixels in the extracted small array that do not overlap with the input array_large. fill_value will be changed to have the same dtype as the array_large array, with one exception. If array_large has integer type and fill_value is np.nan, then a ValueError will be raised.

return_positionbool, optional

If True, return the coordinates of position in the coordinate system of the returned array.


The extracted array.


If return_position is true, this tuple will contain the coordinates of the input position in the coordinate system of array_small. Note that for partially overlapping arrays, new_position might actually be outside of the array_small; array_small[new_position] might give wrong results if any element in new_position is negative.


We consider a large array with the shape 11x10, from which we extract a small array of shape 3x5:

>>> import numpy as np
>>> from astropy.nddata.utils import extract_array
>>> large_array = np.arange(110).reshape((11, 10))
>>> extract_array(large_array, (3, 5), (7, 7))
array([[65, 66, 67, 68, 69],
       [75, 76, 77, 78, 79],
       [85, 86, 87, 88, 89]])