spyrit.core.prep.DirectPoisson.forward

DirectPoisson.forward(x: tensor) → tensor[source]

Preprocess measurements to compensate for the affine image normalization

It computes \(\frac{2}{\alpha}x - H1\), where H1 represents the all-ones vector.

Args:

x: batch of measurement vectors

Shape:

x: \((B, M)\) where \(B\) is the batch dimension

meas_op: the number of measurements meas_op.M should match \(M\).

Output: \((B, M)\)

Example:

>>> x = torch.rand([10,400], dtype=torch.float)
>>> H = torch.rand([400,32*32])
>>> meas_op =  Linear(H)
>>> prep_op = DirectPoisson(1.0, meas_op)
>>> m = prep_op(x)
>>> print(m.shape)
torch.Size([10, 400])