scico.linop.xray.astra¶

X-ray transform LinearOperators wrapping the ASTRA toolbox.

X-ray transform LinearOperator wrapping the parallel beam projections in the ASTRA toolbox. This package provides both C and CUDA implementations of core functionality, but note that use of the CUDA/GPU implementation is expected to result in GPU-host-GPU memory copies when transferring JAX arrays. Other JAX features such as automatic differentiation are not available.

Functions

`angle_to_vector`(det_spacing, angles)	Convert det_spacing and angles to vector geometry specification.
`set_astra_gpu_index`(idx)	Set the index/indices of GPU(s) to be used by astra.

Classes

`XRayTransform2D`(input_shape, det_count, ...)	2D parallel beam X-ray transform based on the ASTRA toolbox.
`XRayTransform3D`(input_shape, det_count[, ...])	3D parallel beam X-ray transform based on the ASTRA toolbox.

scico.linop.xray.astra.set_astra_gpu_index(idx)[source]¶

Set the index/indices of GPU(s) to be used by astra.

Parameters:: idx (Union[int, Sequence[int]]) – Index or indices of GPU(s).

class scico.linop.xray.astra.XRayTransform2D(input_shape, det_count, det_spacing, angles, volume_geometry=None, device='auto')[source]¶

Bases: LinearOperator

2D parallel beam X-ray transform based on the ASTRA toolbox.

Perform tomographic projection (also called X-ray projection) of an image at specified angles, using the ASTRA toolbox.

Parameters:

input_shape (Tuple[int, ...]) – Shape of the input array.
det_spacing (float) – Spacing between detector elements. See the astra documentation for more information..
det_count (int) – Number of detector elements. See the astra documentation for more information.
angles (ndarray) – Array of projection angles in radians.
volume_geometry (Optional[List[float]]) – Specification of the shape of the discretized reconstruction volume. Must either None, in which case it is inferred from input_shape, or follow the syntax described in the astra documentation.
device (str) – Specifies device for projection operation. One of [“auto”, “gpu”, “cpu”]. If “auto”, a GPU is used if available, otherwise, the CPU is used.

fbp(sino, filter_type='Ram-Lak')[source]¶

Filtered back projection (FBP) reconstruction.

Perform tomographic reconstruction using the filtered back projection (FBP) algorithm.

Parameters:

sino (Array) – Sinogram to reconstruct.
filter_type (str) – Select the filter to use. For a list of options see cfg.FilterType in the ASTRA documentation.

Return type:

Array

class scico.linop.xray.astra.XRayTransform3D(input_shape, det_count, det_spacing=None, angles=None, vectors=None)[source]¶

Bases: LinearOperator

3D parallel beam X-ray transform based on the ASTRA toolbox.

Perform tomographic projection (also called X-ray projection) of a volume at specified angles, using the ASTRA toolbox.

This class supports both “parallel3d” and “parallel3d_vec” astra projection geometries. Keyword arguments det_spacing and angles should be specified to use the former, and keyword argument vectors should be specified to use the latter. These options are mutually exclusive.

Parameters:

input_shape (Tuple[int, ...]) – Shape of the input array.
det_count (Tuple[int, int]) – Number of detector elements. See the astra documentation for more information.
det_spacing (Optional[Tuple[float, float]]) – Spacing between detector elements. See the astra documentation for more information.
angles (Optional[ndarray]) – Array of projection angles in radians.
vectors (Optional[ndarray]) – Array of geometry specification vectors.

scico.linop.xray.astra.angle_to_vector(det_spacing, angles)[source]¶

Convert det_spacing and angles to vector geometry specification.

Parameters:

det_spacing (Tuple[float, float]) – Spacing between detector elements. See the astra documentation for more information.
angles (ndarray) – Array of projection angles in radians.

Return type:

ndarray

Returns:

Array of geometry specification vectors.