Resampling

Resampling utilities.

nitransforms.resampling.SERIALIZE_VOLUME_WINDOW_WIDTH: int = 8

Minimum number of volumes to automatically serialize 4D transforms.

nitransforms.resampling.apply(transform: TransformBase, spatialimage: str | Path | SpatialImage, reference: str | Path | SpatialImage = None, order: int = 3, mode: str = 'constant', cval: float = 0.0, prefilter: bool = True, output_dtype: dtype = None, dtype_width: int = 8, serialize_nvols: int = 8, max_concurrent: int = 2) → SpatialImage | ndarray[source]

Apply a transformation to an image, resampling on the reference spatial object.

Parameters:

  • transform (TransformBase) – The 3D, 3D+t, or 4D transform through which data will be resampled.

  • spatialimage (SpatialImage or os.PathLike) – The image object containing the data to be resampled in reference space

  • reference (SpatialImage or os.PathLike) – The image, surface, or combination thereof containing the coordinates of samples that will be sampled.

  • order (int, optional) – The order of the spline interpolation, default is 3. The order has to be in the range 0-5.

  • mode (str, optional) – Determines how the input image is extended when the resamplings overflows a border. One of 'constant', 'reflect', 'nearest', 'mirror', or 'wrap'. Default is 'constant'.

  • cval (float, optional) – Constant value for mode='constant'. Default is 0.0.

  • prefilter (bool, optional) – Determines if the image’s data array is prefiltered with a spline filter before interpolation. The default is True, which will create a temporary float64 array of filtered values if order > 1. If setting this to False, the output will be slightly blurred if order > 1, unless the input is prefiltered, i.e. it is the result of calling the spline filter on the original input.

  • output_dtype (dtype, optional) – The dtype of the returned array or image, if specified. If None, the default behavior is to use the effective dtype of the input image. If slope and/or intercept are defined, the effective dtype is float64, otherwise it is equivalent to the input image’s get_data_dtype() (on-disk type). If reference is defined, then the return value is an image, with a data array of the effective dtype but with the on-disk dtype set to the input image’s on-disk dtype.

  • dtype_width (int) – Cap the width of the input data type to the given number of bytes. This argument is intended to work as a way to implement lower memory requirements in resampling.

  • serialize_nvols (int) – Minimum number of volumes in a 3D+t (that is, a series of 3D transformations independent in time) to resample on a one-by-one basis. Serialized resampling can be executed concurrently (parallelized) with the argument max_concurrent.

  • max_concurrent (int) – Maximum number of 3D resamplings to be executed concurrently.

Returns:

resampled – The data imaged after resampling to reference space.

Return type:

SpatialImage or ndarray

nitransforms.resampling.cap_dtype(dt, nbytes)[source]

Cap the datatype size to shave off memory requirements.

Examples

>>> cap_dtype(np.dtype('f8'), 4)
dtype('float32')

>>> cap_dtype(np.dtype('f8'), 16)
dtype('float64')

>>> cap_dtype('float64', 4)
dtype('float32')

>>> cap_dtype(np.dtype('i1'), 4)
dtype('int8')

>>> cap_dtype('int8', 4)
dtype('int8')

>>> cap_dtype('int32', 1)
dtype('int8')

>>> cap_dtype(np.dtype('i8'), 4)
dtype('int32')