dival.util.odl_utility module

Provides utilities related to ODL.

dival.util.odl_utility.uniform_discr_element(inp, space=None)

Generate an element of a ODL space from an array-like.

Parameters:

• inp (array-like) – The input data from which the element is generated.

• space (odl.discr.DiscretizedSpace, optional) – The space which the element will belong to. If not given, a uniform discretization space with cell size 1 centered around the origin is generated.

dival.util.odl_utility.apply_noise(x, noise_type, noise_kwargs=None, seed=None, random_state=None)

Apply noise to an odl element.

Calls noise functions from odl.phantom.noise or their equivalents from dival.util.odl_noise_random_state.

Parameters:

• x (odl element) – The element to which the noise is applied (in-place).

• noise_type ({'white', 'uniform', 'poisson', 'salt_pepper'}) – Type of noise.

• noise_kwargs (dict, optional) –

  Keyword arguments to be passed to the noise function, e.g. 'stddev' for 'white' noise. The arguments are:

  • for noise_type='white':

    • 'stddev': float, optional

      Standard deviation of each component of the normal distribution. Default is 1.

    • 'relative_stddev': bool, optional

      Whether to multiply 'stddev' with mean(abs(x)). Default is False.

  • for noise_type='poisson':

    • 'scaling_factor': float, optional

      If specified, the intensity is multiplied and the samples from the poisson distribution are divided by this factor: poisson(x * scaling_factor) / scaling_factor. Default is None.

• seed (int, optional) – Random seed passed to the noise function.

• random_state (np.random.RandomState, optional) – Random state passed to the noise function.

class dival.util.odl_utility.NoiseOperator(*args, **kwargs)

Bases: Operator

Operator applying noise.

Wraps apply_noise(), which calls noise functions from odl.phantom.noise or their equivalents from dival.util.odl_noise_random_state.

__init__(domain, noise_type, noise_kwargs=None, seed=None, random_state=None)

Parameters:

• space (odl space) – Domain and range.

• noise_type ({'white', 'uniform', 'poisson', 'salt_pepper'}) – Type of noise.

• noise_kwargs (dict, optional) – Keyword arguments to be passed to the noise function, cf. docs for apply_noise().

• seed (int, optional) – Random seed passed to the noise function.

• random_state (np.random.RandomState, optional) – Random state passed to the noise function.

class dival.util.odl_utility.CallbackStore(results=None, function=None, step=1)

Bases: Callback

This is a modified copy of odl.solvers.util.callback.CallbackStore, Copyright held by The ODL contributors, subject to the terms of the Mozilla Public License, v. 2.0. If a copy of the MPL was not distributed with this file, You can obtain one at https://mozilla.org/MPL/2.0/. This copy incorporates https://github.com/odlgroup/odl/pull/1539.

Callback for storing all iterates of a solver. Can optionally apply a function, for example the norm or calculating the residual. By default, calls the copy() method on the iterates before storing.

__init__(results=None, function=None, step=1)

Initialize a new instance.

Parameters:

• results (list, optional) – List in which to store the iterates. Default: new list ([])

• function (callable, optional) – Deprecated, use composition instead. See examples. Function to be called on all incoming results before storage. Default: copy

• step (int, optional) – Number of iterates between storing iterates.

Examples

Store results as-is: >>> callback = CallbackStore() Provide list to store iterates in: >>> results = [] >>> callback = CallbackStore(results=results) Store the norm of the results: >>> norm_function = lambda x: x.norm() >>> callback = CallbackStore() * norm_function

reset()

Clear the results list.

class dival.util.odl_utility.CallbackStoreAfter(results=None, store_after_iters=None)

Bases: Callback

Callback for storing after specific numbers of iterations of a solver. Calls the copy() method on the iterates before storing.

The source code of this class is based on odl.solvers.util.callback.CallbackStore, Copyright held by The ODL contributors, subject to the terms of the Mozilla Public License, v. 2.0. If a copy of the MPL was not distributed with this file, You can obtain one at https://mozilla.org/MPL/2.0/.

__init__(results=None, store_after_iters=None)

Initialize a new instance.

Parameters:

• results (list, optional) – List in which to store the iterates. Default: new list ([])

• store_after_iters (list of int, optional) – Numbers of iterations after which the result should be stored.

reset()

Clear the results list.

class dival.util.odl_utility.ResizeOperator(*args, **kwargs)

Bases: Operator

__init__(reco_space, space, order=1)

Initialize a new instance.

Parameters:

• domain (Set) – The domain of this operator, i.e., the set of elements to which this operator can be applied

• range (Set) – The range of this operator, i.e., the set this operator maps to

• linear (bool, optional) – If True, the operator is considered as linear. In this case, domain and range have to be instances of LinearSpace, or Field.

class dival.util.odl_utility.RayBackProjection(*args, **kwargs)

Bases: Operator

Adjoint of the discrete Ray transform between L^p spaces.

This class is copied and modified from odl.

This main-scope class definition is used by patch_ray_trafo_for_pickling() to make a ray transform object pickleable by replacing its _adjoint attribute with an instance of this class.

__init__(ray_trafo, **kwargs)

Initialize a new instance.

Parameters:

• domain (Set) – The domain of this operator, i.e., the set of elements to which this operator can be applied

• range (Set) – The range of this operator, i.e., the set this operator maps to

• linear (bool, optional) – If True, the operator is considered as linear. In this case, domain and range have to be instances of LinearSpace, or Field.

property geometry

property adjoint

Adjoint of this operator (abstract).

Raises:

OpNotImplementedError – Since the adjoint cannot be default implemented.

dival.util.odl_utility.patch_ray_trafo_for_pickling(ray_trafo)

Make an object of type odl.tomo.operators.RayTransform pickleable by overwriting the _adjoint (which originally has a local class type) with a dival.util.torch_utility.RayBackProjection object. This can be required for multiprocessing.

Parameters:

ray_trafo (odl.tomo.operators.RayTransform) – The ray transform to patch for pickling.