scipy.ndimage.

label#

scipy.ndimage.label(input, structure=None, output=None)[source]#

Label features in an array.

Parameters:

inputarray_like

An array-like object to be labeled. Any non-zero values in input are counted as features and zero values are considered the background.

structurearray_like, optional

A structuring element that defines feature connections. structure must be centrosymmetric (see Notes). If no structuring element is provided, one is automatically generated with a squared connectivity equal to one. That is, for a 2-D input array, the default structuring element is:

[[0,1,0],
 [1,1,1],
 [0,1,0]]

output(None, data-type, array_like), optional

If output is a data type, it specifies the type of the resulting labeled feature array. If output is an array-like object, then output will be updated with the labeled features from this function. This function can operate in-place, by passing output=input. Note that the output must be able to store the largest label, or this function will raise an Exception.

Returns:

labelndarray or int

An integer ndarray where each unique feature in input has a unique label in the returned array.

num_featuresint

How many objects were found.

If output is None, this function returns a tuple of (labeled_array, num_features).

If output is a ndarray, then it will be updated with values in labeled_array and only num_features will be returned by this function.

See also

find_objects: generate a list of slices for the labeled features (or objects); useful for finding features’ position or dimensions

Notes

A centrosymmetric matrix is a matrix that is symmetric about the center. See [1] for more information.

The structure matrix must be centrosymmetric to ensure two-way connections. For instance, if the structure matrix is not centrosymmetric and is defined as:

[[0,1,0],
 [1,1,0],
 [0,0,0]]

and the input is:

[[1,2],
 [0,3]]

then the structure matrix would indicate the entry 2 in the input is connected to 1, but 1 is not connected to 2.

label has experimental support for Python Array API Standard compatible backends in addition to NumPy. Please consider testing these features by setting an environment variable SCIPY_ARRAY_API=1 and providing CuPy, PyTorch, JAX, or Dask arrays as array arguments. The following combinations of backend and device (or other capability) are supported.

Library	CPU	GPU
NumPy	✅	n/a
CuPy	n/a	✅
PyTorch	✅	⛔
JAX	⚠️ no JIT	⛔
Dask	⚠️ computes graph	n/a

See Support for the array API standard for more information.

References

[1]

James R. Weaver, “Centrosymmetric (cross-symmetric) matrices, their basic properties, eigenvalues, and eigenvectors.” The American Mathematical Monthly 92.10 (1985): 711-717.

Examples

Create an image with some features, then label it using the default (cross-shaped) structuring element:

>>> from scipy.ndimage import label, generate_binary_structure
>>> import numpy as np
>>> a = np.array([[0,0,1,1,0,0],
...               [0,0,0,1,0,0],
...               [1,1,0,0,1,0],
...               [0,0,0,1,0,0]])
>>> labeled_array, num_features = label(a)

Each of the 4 features are labeled with a different integer:

>>> num_features
4
>>> labeled_array
array([[0, 0, 1, 1, 0, 0],
       [0, 0, 0, 1, 0, 0],
       [2, 2, 0, 0, 3, 0],
       [0, 0, 0, 4, 0, 0]], dtype=int32)

Generate a structuring element that will consider features connected even if they touch diagonally:

>>> s = generate_binary_structure(2,2)

or,

>>> s = [[1,1,1],
...      [1,1,1],
...      [1,1,1]]

Label the image using the new structuring element:

>>> labeled_array, num_features = label(a, structure=s)

Show the 2 labeled features (note that features 1, 3, and 4 from above are now considered a single feature):

>>> num_features
2
>>> labeled_array
array([[0, 0, 1, 1, 0, 0],
       [0, 0, 0, 1, 0, 0],
       [2, 2, 0, 0, 1, 0],
       [0, 0, 0, 1, 0, 0]], dtype=int32)