Class FillHoles

monai/transforms/post/array.py:511–586

r""" This transform fills holes in the image and can be used to remove artifacts inside segments. An enclosed hole is defined as a background pixel/voxel which is only enclosed by a single class. The definition of enclosed can be defined with the connectivity parameter:: 1-conn

class FillHoles(Transform):
    r"""
    This transform fills holes in the image and can be used to remove artifacts inside segments.

    An enclosed hole is defined as a background pixel/voxel which is only enclosed by a single class.
    The definition of enclosed can be defined with the connectivity parameter::

        1-connectivity     2-connectivity     diagonal connection close-up

             [ ]           [ ]  [ ]  [ ]             [ ]
              |               \  |  /                 |  <- hop 2
        [ ]--[x]--[ ]      [ ]--[x]--[ ]        [x]--[ ]
              |               /  |  \             hop 1
             [ ]           [ ]  [ ]  [ ]

    It is possible to define for which labels the hole filling should be applied.
    The input image is assumed to be a PyTorch Tensor or numpy array with shape [C, spatial_dim1[, spatial_dim2, ...]].
    If C = 1, then the values correspond to expected labels.
    If C > 1, then a one-hot-encoding is expected where the index of C matches the label indexing.

    Note:

        The label 0 will be treated as background and the enclosed holes will be set to the neighboring class label.

        The performance of this method heavily depends on the number of labels.
        It is a bit faster if the list of `applied_labels` is provided.
        Limiting the number of `applied_labels` results in a big decrease in processing time.

    For example:

        Use FillHoles with default parameters::

            [1, 1, 1, 2, 2, 2, 3, 3]         [1, 1, 1, 2, 2, 2, 3, 3]
            [1, 0, 1, 2, 0, 0, 3, 0]    =>   [1, 1 ,1, 2, 0, 0, 3, 0]
            [1, 1, 1, 2, 2, 2, 3, 3]         [1, 1, 1, 2, 2, 2, 3, 3]

        The hole in label 1 is fully enclosed and therefore filled with label 1.
        The background label near label 2 and 3 is not fully enclosed and therefore not filled.
    """

    backend = [TransformBackends.NUMPY]

    def __init__(self, applied_labels: Iterable[int] | int | None = None, connectivity: int | None = None) -> None:
        """
        Initialize the connectivity and limit the labels for which holes are filled.

        Args:
            applied_labels: Labels for which to fill holes. Defaults to None, that is filling holes for all labels.
            connectivity: Maximum number of orthogonal hops to consider a pixel/voxel as a neighbor.
                Accepted values are ranging from  1 to input.ndim. Defaults to a full connectivity of ``input.ndim``.
        """
        super().__init__()
        self.applied_labels = ensure_tuple(applied_labels) if applied_labels else None
        self.connectivity = connectivity

    def __call__(self, img: NdarrayOrTensor) -> NdarrayOrTensor:
        """
        Fill the holes in the provided image.