Radon Projections

Parallel Beam

class torch_radon.Radon(resolution: int, angles, det_count=- 1, det_spacing=1.0, clip_to_circle=False)

Class that implements Radon projection for the Parallel Beam geometry.

Parameters

• resolution – The resolution of the input images.

• angles – Array containing the list of measuring angles. Can be a Numpy array or a PyTorch tensor.

• det_count – Number of rays that will be projected. By default it is = resolution

• det_spacing – Distance between two contiguous rays.

• clip_to_circle – If True both forward and backward projection will be restricted to pixels inside the circle (highlighted in cyan).

Note

Currently only support resolutions which are multiples of 16.

forward(self, x)

Radon forward projection.

Parameters

x – PyTorch GPU tensor with shape \((d_1, \dots, d_n, r, r)\) where \(r\) is the resolution given to the constructor of this class.

Returns

PyTorch GPU tensor containing sinograms. Has shape \((d_1, \dots, d_n, len(angles), det\_count)\).

backprojection(self, sinogram)

Radon backward projection.

Parameters

sinogram – PyTorch GPU tensor containing sinograms with shape \((d_1, \dots, d_n, len(angles), det\_count)\).

Returns

PyTorch GPU tensor with shape \((d_1, \dots, d_n, r, r)\) where \(r\) is the resolution given to the constructor of this class.

backward(self, sinogram)

Same as backprojection

Fanbeam

class torch_radon.RadonFanbeam(resolution: int, angles, source_distance: float, det_distance: float = - 1, det_count: int = - 1, det_spacing: float = - 1, clip_to_circle=False)

Class that implements Radon projection for the Fanbeam geometry.

Parameters

• resolution – The resolution of the input images.

• angles – Array containing the list of measuring angles. Can be a Numpy array or a PyTorch tensor.

• source_distance – Distance between the source of rays and the center of the image.

• det_distance – Distance between the detector plane and the center of the image. By default it is = source_distance.

• det_count – Number of rays that will be projected. By default it is = resolution.

• det_spacing – Distance between two contiguous rays.

• clip_to_circle – If True both forward and backward projection will be restricted to pixels inside the circle (highlighted in cyan).

Note

Currently only support resolutions which are multiples of 16.

forward(self, x)

Radon forward projection.

Parameters

x – PyTorch GPU tensor with shape \((d_1, \dots, d_n, r, r)\) where \(r\) is the resolution given to the constructor of this class.

Returns

PyTorch GPU tensor containing sinograms. Has shape \((d_1, \dots, d_n, len(angles), det\_count)\).

backprojection(self, sinogram)

Radon backward projection.

Parameters

sinogram – PyTorch GPU tensor containing sinograms with shape \((d_1, \dots, d_n, len(angles), det\_count)\).

Returns

PyTorch GPU tensor with shape \((d_1, \dots, d_n, r, r)\) where \(r\) is the resolution given to the constructor of this class.

backward(self, sinogram)

Same as backprojection