PnP.py¶
Source: LION/reconstructors/PnP.py
Plug-and-Play (PnP) Reconstructor using a prior function.
- class LION.reconstructors.PnP.PnP(physics, prior_fn, algorithm='ADMM')[source]¶
Bases:
LIONReconstructor- Parameters:
physics (Geometry | Operator)
prior_fn (Callable[[torch.Tensor], torch.Tensor])
algorithm (Literal['ADMM', 'HQS', 'FBS'])
- static cite(cite_format='MLA')[source]¶
Print the Plug-and-Play ADMM citation.
- Parameters:
cite_format (str)
- Return type:
None
- reconstruct_sample(sino, *, prog_bar=False, **kwargs)[source]¶
Reconstruct the sinogram using the model and geometry.
- Parameters:
sino – Sinogram tensor.
prog_bar (bool)
- Returns:
Reconstructed image tensor.
- hqs_algorithm(sino, *, lambda_=0.23, mu=0.1, max_iter=100, noise_level=None, prog_bar=False)[source]¶
Placeholder for the Half Quadratic Splitting algorithm implementation.
- Parameters:
sino – Sinogram tensor.
lambda – Regularization parameter.
mu – Step size.
max_iter – Maximum number of iterations.
prog_bar (bool)
- Returns:
Reconstructed image tensor.