Morphology Utils

Boolean-mask morphology and region helpers.

These reproduce the OpenCV morphology operations the analysis pipelines used to call, so mask post-processing stays available without the opencv-python dependency. Dilation is pure NumPy; region labeling uses scikit-image.

`dilate_rect(mask, size)`

Dilate a boolean mask with a rectangular all-ones kernel.

A rectangular kernel is separable, so the mask is dilated along columns and then along rows, each in one running-sum pass independent of the kernel extent. The result matches OpenCV's rectangular dilation with constant zero-padding. Only odd kernel extents are supported so the anchor is unambiguous.

Parameters:

Name	Type	Description	Default
`mask`	`ndarray`	Two-dimensional boolean mask to dilate.	required
`size`	`tuple[int, int]`	Kernel `(width, height)`, both odd.	required

Returns:

Type	Description
`ndarray`	np.ndarray: Dilated boolean mask, same shape as `mask`.

Raises:

Type	Description
`ValueError`	If `mask` is not two-dimensional, or either kernel extent is even, not positive, or greater than 65535.

`mask_regions(mask, min_area=1)`

Return bounding boxes of the connected foreground regions in a boolean mask.

Delegates labeling to skimage.measure.label with full connectivity, so pixels touching along an edge or a corner belong to the same region.

Parameters:

Name	Type	Description	Default
`mask`	`ndarray`	Two-dimensional boolean mask.	required
`min_area`	`int`	Minimum number of foreground pixels a region needs to be kept. Defaults to 1.	`1`

Returns:

Type	Description
`list[tuple[int, int, int, int]]`	list[tuple[int, int, int, int]]: `(x, y, width, height)` boxes ordered top to bottom, then left to right.

Raises:

Type	Description
`ImportError`	If scikit-image is unavailable.
`ValueError`	If `mask` is not two-dimensional or `min_area` is less than 1.