`skimage.segmentation._join:relabel_sequential`

source: /dev/scikit-image/src/skimage/segmentation/_join.py :76

Signature

def relabel_sequential ( label_field , offset = 1 )

Summary

Relabel arbitrary labels to {offset, ... offset + number_of_labels}.

Extended Summary

This function also returns the forward map (mapping the original labels to the reduced labels) and the inverse map (mapping the reduced labels back to the original ones).

Parameters

label_field : numpy array of int, arbitrary shape: An array of labels, which must be non-negative integers.
offset : int, optional: The return labels will start at offset, which should be strictly positive.

Returns

relabeled : numpy array of int, same shape as `label_field`: The input label field with labels mapped to {offset, ..., number_of_labels + offset - 1}. The data type will be the same as label_field, except when offset + number_of_labels causes overflow of the current data type.
forward_map : ArrayMap: The map from the original label space to the returned label space. Can be used to re-apply the same mapping. See examples for usage. The output data type will be the same as relabeled.
inverse_map : ArrayMap: The map from the new label space to the original space. This can be used to reconstruct the original label field from the relabeled one. The output data type will be the same as label_field.

Notes

The label 0 is assumed to denote the background and is never remapped.

The forward map can be extremely big for some inputs, since its length is given by the maximum of the label field. However, in most situations, label_field.max() is much smaller than label_field.size, and in these cases the forward map is guaranteed to be smaller than either the input or output images.

Examples

from skimage.segmentation import relabel_sequential
label_field = np.array([1, 1, 5, 5, 8, 99, 42])
relab, fw, inv = relabel_sequential(label_field)
relab
print(fw)
np.array(fw)
np.array(inv)

✓

(fw[label_field] == relab).all()
(inv[relab] == label_field).all()

✗

relab, fw, inv = relabel_sequential(label_field, offset=5)
relab