apps.partial_fourier

bartorch.apps.partial_fourier()

Homodyne completion of an image reconstructed from partial-Fourier k-space.

An axis is partial Fourier when the lines missing at one end of sampled outnumber the widest gap between sampled lines, which undersampling alone leaves. Each such axis is completed by bart homodyne -I -C with the acquired fraction of its lines, on the flipped image when the acquired side is the high indices. Other axes are untouched.

Parameters:

• image (torch.Tensor) – Complex image ([z,] y, x) reconstructed from the sampled k-space. The Fourier transform relating the two is centred: the zero frequency is at index n // 2 of each axis.

• sampled (torch.Tensor) – Boolean sampling mask of that k-space, with the shape of image; True where a sample was acquired.

Returns:

Complex image with the shape of image. A completed image is real-valued: homodyne reconstruction removes the phase estimated from the central lines of k-space and keeps the real part.

Return type:

torch.Tensor

Raises:

ValueError – If sampled and image differ in shape.