A Pytorch implementation of the radon operator and filtered backprojection with, except for a constant, adjoint radon operator and backprojection.
- clone repository
- install pytorch: https://pytorch.org/get-started/locally/
- Install Radon-Transformation:
pip install ./Radon-Transformation/
from radon_transformation.radon import get_operators
input = # Load some input of shape (bsz x 1 x w x h)
radon_op, fbp_op = get_operators(n_angles=200, image_size=input.shape[-1], device='cuda')
sino = radon_op(input)
reconstructed = fbp_op(sino)