bbox

Cython code for computing per-label axis-aligned bounding boxes on integer-labelled images.

For each label l present in the input array the result contains the minimum and maximum coordinate along every axis:

result[l, axis, 0]  →  min coordinate
result[l, axis, 1]  →  max coordinate

Functions

bbox_3d()

Bounding boxes for a 3-D label image.

bbox_2d()

Bounding boxes for a 2-D label image.

bbox_1d()

Bounding boxes for a 1-D label image.

Notes

Label 0 is included in the output (background). The caller can discard result[0] if the background bounding box is not needed.

The max + 1 label count may overflow for images whose maximum label value equals the dtype maximum (e.g. 255 for uint8). This is intentional and flagged with an inline warning.

bbox_1d(img_)

Compute per-label bounding boxes for a 1-D label image.

Parameters:

img (ndarray, 1-D, integer dtype) – Label image.

Returns

ndarray, shape (n_labels, 1, 2), dtype uint32

result[l, 0, 0/1] gives the min/max coordinate of label l.

bbox_2d(img_)

Compute per-label bounding boxes for a 2-D label image.

Parameters:

img (ndarray, 2-D, integer dtype) – Label image.

Returns

ndarray, shape (n_labels, 2, 2), dtype uint32

result[l, axis, 0/1] gives the min/max coordinate of label l along axis.

bbox_3d(img_)

Compute per-label bounding boxes for a 3-D label image.

Parameters:

img (ndarray, 3-D, integer dtype) – Label image.

Returns

ndarray, shape (n_labels, 3, 2), dtype uint32

result[l, axis, 0/1] gives the min/max coordinate of label l along axis.