Class ImageFilter

monai/transforms/utility/array.py:1534–1786

Applies a convolution filter to the input image. Args: filter: A string specifying the filter, a custom filter as ``torch.Tenor`` or ``np.ndarray`` or a ``nn.Module``. Available options for string are: ``mean``, ``laplace``, ``elliptical``, ``sobel``, ``shar

class ImageFilter(Transform):
    """
    Applies a convolution filter to the input image.

    Args:
        filter:
            A string specifying the filter, a custom filter as ``torch.Tenor`` or ``np.ndarray`` or a ``nn.Module``.
            Available options for string are: ``mean``, ``laplace``, ``elliptical``, ``sobel``, ``sharpen``, ``median``, ``gauss``
            See below for short explanations on every filter.
        filter_size:
            A single integer value specifying the size of the quadratic or cubic filter.
            Computational complexity scales to the power of 2 (2D filter) or 3 (3D filter), which
            should be considered when choosing filter size.
        kwargs:
            Additional arguments passed to filter function, required by ``sobel`` and ``gauss``.
            See below for details.

    Raises:
        ValueError: When ``filter_size`` is not an uneven integer
        ValueError: When ``filter`` is an array and ``ndim`` is not in [1,2,3]
        ValueError: When ``filter`` is an array and any dimension has an even shape
        NotImplementedError: When ``filter`` is a string and not in ``self.supported_filters``
        KeyError: When necessary ``kwargs`` are not passed to a filter that requires additional arguments.


    **Mean Filtering:** ``filter='mean'``

    Mean filtering can smooth edges and remove aliasing artifacts in an segmentation image.
    See also py:func:`monai.networks.layers.simplelayers.MeanFilter`
    Example 2D filter (5 x 5)::

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

    If smoothing labels with this filter, ensure they are in one-hot format.

    **Outline Detection:** ``filter='laplace'``

    Laplacian filtering for outline detection in images. Can be used to transform labels to contours.
    See also py:func:`monai.networks.layers.simplelayers.LaplaceFilter`

    Example 2D filter (5x5)::

        [[-1., -1., -1., -1., -1.],
         [-1., -1., -1., -1., -1.],
         [-1., -1., 24., -1., -1.],
         [-1., -1., -1., -1., -1.],
         [-1., -1., -1., -1., -1.]]


    **Dilation:** ``filter='elliptical'``

    An elliptical filter can be used to dilate labels or label-contours.
    Example 2D filter (5x5)::