spyrit.core.meas.HadamSplit2d.measure

HadamSplit2d.measure(x: tensor) → tensor[source]

Simulate noiseless measurements from matrix A.

It computes

\[y =\mathcal{S}\left(AXA^T\right),\]

where \(\mathcal{S} \colon\, \mathbb{R}^{2h\times 2h} \to \mathbb{R}^{2M}\) is the subsampling operator, \(A \colon\, \mathbb{R}_+^{2h\times h}\) is the acquisition matrix that contains the positive and negative component of 2D Hadamard patterns, \(X \in \mathbb{R}^{h\times h}\) is the (2D) image, \(2M\) is the number of DMD patterns, and \(h\) is the image size.

Args:

x (torch.tensor): Image \(X\) whose dimensions self.meas_dims must have shape shape self.meas_shape.

Returns:

Measurement vector \(y \in \mathbb{R}^{2M}\).

Examples:

Example 1: No subsampling

>>> import torch
>>> import spyrit.core.meas as meas
>>> h = 32
>>> Ord = torch.randn(h, h)
>>> meas_op = meas.HadamSplit2d(h)
>>> x = torch.empty(10, h, h).uniform_(0, 1)
>>> y = meas_op.measure(x)
>>> print(y.shape)
torch.Size([10, 2048])

Example 2: With subsampling

>>> import torch
>>> import spyrit.core.meas as meas
>>> h = 32
>>> Ord = torch.randn(h, h)
>>> meas_op = meas.HadamSplit2d(h, 49)
>>> x = torch.empty(8, 2, h, h).uniform_(0, 1)
>>> y = meas_op.measure_H(x)
>>> print(y.shape)
torch.Size([8, 2, 49])