spyrit.core.recon.TikhoNet.reconstruct_pinv

TikhoNet.reconstruct_pinv(y)

Reconstruct an image from measurements without denoising.

This method sucessively applies the preprocessing operator prep and the Tikhonov regularization operator tikho to the input measurement vectors x.

Important

The measurements passed as input must NOT be preprocessed.

Args:

y: raw measurement vectors. Have shape \((b, c, m)\)

Returns:

torch.tensor: Reconstructed images. Have shape \((b,c,h,w)\) if reshape_output is True in the kwargs dictionary (default) or \((b,c,hw)\) otherwise.

Example:

>>> import spyrit
>>> acqu = spyrit.core.meas.HadamSplit2d(8)
>>> prep = spyrit.core.prep.UnsplitRescale()
>>> sigma = torch.ones(64, 64)
>>> tikho = TikhoNet(acqu, prep, sigma)
>>> x = torch.rand(10, 1, 8, 8)
>>> y = acqu(x)
>>> z = tikho.reconstruct_pinv(y)
>>> print(z.shape)
torch.Size([10, 1, 8, 8])