correct_global_phase#
- deepinv.optim.phase_retrieval.correct_global_phase(x_recon, x, threshold=1e-5, verbose=False)[source]#
Corrects the global phase of the reconstructed image.
Warning
Do not mix the order of the reconstructed and original images since this function modifies x_recon in place.
The global phase shift is comptued per image and per channel as:
\[e^{-i \phi} = \frac{\conj{\hat{x}} \cdot x}{|x|^2},\]where \(\conj{\hat{x}}\) is the complex conjugate of the reconstructed image, \(x\) is the reference image, and \(|x|^2\) is the squared magnitude of the reference image.
The global phase shift is then applied to the reconstructed image as:
\[\hat{x} = \hat{x} \cdot e^{-i \phi},\]for the corresponding image and channel.
- param torch.Tensor x_recon:
Reconstructed image.
- param torch.Tensor x:
Original image.
- param float threshold:
Threshold to determine if the global phase shift is constant. Default is 1e-5.
- param bool verbose:
If True, prints information about the global phase shift. Default is False.
- return:
The corrected image.
Examples using correct_global_phase:#
Random phase retrieval and reconstruction methods.