spyrit.core.prep.SplitPoisson.forward

SplitPoisson.forward(x: tensor) → tensor

Preprocess to compensates for image normalization and splitting of the measurement operator.

It computes \(\frac{x[0::2]-x[1::2]}{\alpha} - H1\)

Args:

x: batch of measurement vectors

Shape:

x: \((*, 2M)\) where \(*\) indicates one or more dimensions

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

Output: \((*, M)\)

Example:

>>> x = torch.rand([10,2*400], dtype=torch.float)
>>> H = torch.rand([400,32*32])
>>> meas_op =  LinearSplit(H)
>>> split_op = SplitPoisson(10, meas_op)
>>> m = split_op(x)
>>> print(m.shape)
torch.Size([10, 400])

Example 2:

>>> x = torch.rand([10,2*400], dtype=torch.float)
>>> Perm = torch.rand([32,32])
>>> meas_op = HadamSplit(400, 32,  Perm)
>>> split_op = SplitPoisson(10, meas_op)
>>> m = split_op(x)
>>> print(m.shape)
torch.Size([10, 400])