mrpro.utils.filters.filter_separable

mrpro.utils.filters.filter_separable(x: Tensor, kernels: Sequence[Tensor], dim: Sequence[int], pad_mode: Literal['constant', 'reflect', 'replicate', 'circular', 'none'] = 'constant', pad_value: float = 0.0) → Tensor

Apply the separable filter kernels to the tensor x along the axes dim.

Does padding to keep the output the same size as the input.

Parameters:

• x – Tensor to filter

• kernels – List of 1D kernels to apply to the tensor x

• dim – Axes to filter over. Must have the same length as kernels.

• pad_mode – Padding mode

• pad_value – Padding value for pad_mode = constant

Returns:

• The filtered tensor, with the same shape as the input unless pad_mode is ‘none’ and

• and promoted dtype of the input and the kernels.