def phase_retrieval(args):
"""
Perform single-material phase retrieval
using projection data.
Parameters
----------
data : ndarray
Projection data.
pixel_size : scalar
Detector pixel size in cm.
dist : scalar
Propagation distance of x-rays in cm.
energy : scalar
Energy of x-rays in keV.
alpha : scalar, optional
Regularization parameter.
padding : bool, optional
Applies padding for Fourier transform. For quick testing
you can use False for faster results.
Returns
-------
phase : ndarray
Retrieved phase.
References
----------
- `J. of Microscopy, Vol 206(1), 33-40, 2001 \
<http://onlinelibrary.wiley.com/doi/10.1046/j.1365-2818.2002.01010.x/abstract>`_
"""
data, args, ind_start, ind_end = args
H, x_shift, y_shift, tmp_proj, padding = args
num_proj, dx, dy = data.shape # dx:slices, dy:pixels
for m in range(num_proj):
proj = data[m, :, :]
if padding:
tmp_proj[x_shift:dx+x_shift, y_shift:dy+y_shift] = proj
fft_proj = fftw.fftw2(tmp_proj)
filtered_proj = np.multiply(H, fft_proj)
tmp = np.real(fftw.ifftw2(filtered_proj))
proj = tmp[x_shift:dx+x_shift, y_shift:dy+y_shift]
elif not padding:
fft_proj = fftw.fftw2(proj)
filtered_proj = np.multiply(H, fft_proj)
proj = np.real(fftw.ifftw2(filtered_proj)) / np.max(H)
data[m, :, :] = proj
return ind_start, ind_end, data
def upsample2df(data, level):
"""
Upsample the slices in Fourier domain.
Parameters
----------
data : ndarray, float32
3-D reconstructed data with dimensions:
[slices, pixels, pixels]
level : scalar, int32
Upsampling level. For example level=2
means, the sinogram will be upsampled by 4,
and level=3 means upsampled by 8.
Returns
-------
output : ndarray
Downsampled reconstructed 3-D data with dimensions:
[slices, pixels*level^2, pixels*level^2]
"""
num_slices = np.array(data.shape[0], dtype='int32')
num_pixels = np.array(data.shape[1], dtype='int32')
binsize = np.power(2, level)
fftw2data = np.zeros((num_pixels*binsize,
num_pixels*binsize),
dtype='complex')
upsampled_data = np.zeros((num_slices,
num_pixels*binsize,
num_pixels*binsize),
dtype='float32')
ind = slice(num_pixels*(binsize-1)/2, num_pixels*(binsize-1)/2+num_pixels, 1)
for m in range(num_slices):
fftw2data[ind, ind] = np.fft.fftshift(fftw2(data[m, :, :]))
upsampled_data[m, :, :] = np.real(ifftw2(np.fft.ifftshift(fftw2data)))
return upsampled_data
def upsample2df(data, level):
"""
Upsample the slices in Fourier domain.
Parameters
----------
data : ndarray, float32
3-D reconstructed data with dimensions:
[slices, pixels, pixels]
level : scalar, int32
Upsampling level. For example level=2
means, the sinogram will be upsampled by 4,
and level=3 means upsampled by 8.
Returns
-------
output : ndarray
Downsampled reconstructed 3-D data with dimensions:
[slices, pixels*level^2, pixels*level^2]
"""
num_slices = np.array(data.shape[0], dtype='int32')
num_pixels = np.array(data.shape[1], dtype='int32')
binsize = np.power(2, level)
fftw2data = np.zeros((num_pixels * binsize, num_pixels * binsize),
dtype='complex')
upsampled_data = np.zeros(
(num_slices, num_pixels * binsize, num_pixels * binsize),
dtype='float32')
ind = slice(num_pixels * (binsize - 1) / 2,
num_pixels * (binsize - 1) / 2 + num_pixels, 1)
for m in range(num_slices):
fftw2data[ind, ind] = np.fft.fftshift(fftw2(data[m, :, :]))
upsampled_data[m, :, :] = np.real(ifftw2(np.fft.ifftshift(fftw2data)))
return upsampled_data
def phase_retrieval(args):
"""
Perform single-material phase retrieval
using projection data.
Parameters
----------
data : ndarray
3-D tomographic data with dimensions:
[projections, slices, pixels]
pixel_size : scalar
Detector pixel size in cm.
dist : scalar
Propagation distance of x-rays in cm.
energy : scalar
Energy of x-rays in keV.
alpha : scalar, optional
Regularization parameter.
padding : bool, optional
Applies padding for Fourier transform. For quick testing
you can use False for faster results.
Returns
-------
phase : ndarray
Retrieved phase.
References
----------
- `J. of Microscopy, Vol 206(1), 33-40, 2001 \
<http://onlinelibrary.wiley.com/doi/10.1046/j.1365-2818.2002.01010.x/abstract>`_
Examples
--------
- Phase retrieval:
>>> import tomopy
>>>
>>> # Load data
>>> myfile = 'demo/data.h5'
>>> data, white, dark, theta = tomopy.xtomo_reader(myfile)
>>>
>>> # Construct tomo object
>>> d = tomopy.xtomo_dataset(log='error')
>>> d.dataset(data, white, dark, theta)
>>> d.normalize()
>>>
>>> # Reconstruct data before phase retrieval
>>> d.center=661.5
>>> d.gridrec()
>>>
>>> # Save reconstructed data before phase retrieval
>>> output_file='tmp/before_phase_retrieval_'
>>> tomopy.xtomo_writer(d.data_recon, output_file)
>>> print "Images are succesfully saved at " + output_file + '...'
>>>
>>> # Reconstruct data after phase retrieval
>>> d.phase_retrieval(pixel_size=1e-4, dist=70, energy=20)
>>> d.gridrec()
>>>
>>> # Save reconstructed data after phase retrieval
>>> output_file='tmp/after_phase_retrieval_'
>>> tomopy.xtomo_writer(d.data_recon, output_file)
>>> print "Images are succesfully saved at " + output_file + '...'
"""
# Arguments passed by multi-processing wrapper
ind, dshape, inputs = args
# Function inputs
data = mp.tonumpyarray(mp.shared_arr, dshape) # shared-array
pixel_size, dist, energy, alpha, padding = inputs
# Compute the filter.
H, x_shift, y_shift, tmp_proj = paganin_filter(data,
pixel_size, dist, energy, alpha, padding)
num_proj, dx, dy = dshape # dx:slices, dy:pixels
for m in ind:
proj = data[m, :, :]
if padding:
tmp_proj[x_shift:dx+x_shift, y_shift:dy+y_shift] = proj
fft_proj = fftw.fftw2(tmp_proj)
filtered_proj = np.multiply(H, fft_proj)
tmp = np.real(fftw.ifftw2(filtered_proj))/np.max(H)
proj = tmp[x_shift:dx+x_shift, y_shift:dy+y_shift]
elif not padding:
fft_proj = fftw.fftw2(proj)
filtered_proj = np.multiply(H, fft_proj)
proj = np.real(fftw.ifftw2(filtered_proj))/np.max(H)
data[m, :, :] = proj
def phase_retrieval(args):
"""
Perform single-material phase retrieval
using projection data.
Parameters
----------
data : ndarray
3-D tomographic data with dimensions:
[projections, slices, pixels]
pixel_size : scalar
Detector pixel size in cm.
dist : scalar
Propagation distance of x-rays in cm.
energy : scalar
Energy of x-rays in keV.
alpha : scalar, optional
Regularization parameter.
padding : bool, optional
Applies padding for Fourier transform. For quick testing
you can use False for faster results.
Returns
-------
phase : ndarray
Retrieved phase.
References
----------
- `J. of Microscopy, Vol 206(1), 33-40, 2001 \
<http://onlinelibrary.wiley.com/doi/10.1046/j.1365-2818.2002.01010.x/abstract>`_
Examples
--------
- Phase retrieval:
>>> import tomopy
>>>
>>> # Load data
>>> myfile = 'demo/data.h5'
>>> data, white, dark, theta = tomopy.xtomo_reader(myfile)
>>>
>>> # Construct tomo object
>>> d = tomopy.xtomo_dataset(log='error')
>>> d.dataset(data, white, dark, theta)
>>> d.normalize()
>>>
>>> # Reconstruct data before phase retrieval
>>> d.center=661.5
>>> d.gridrec()
>>>
>>> # Save reconstructed data before phase retrieval
>>> output_file='tmp/before_phase_retrieval_'
>>> tomopy.xtomo_writer(d.data_recon, output_file)
>>> print "Images are succesfully saved at " + output_file + '...'
>>>
>>> # Reconstruct data after phase retrieval
>>> d.phase_retrieval(pixel_size=1e-4, dist=70, energy=20)
>>> d.gridrec()
>>>
>>> # Save reconstructed data after phase retrieval
>>> output_file='tmp/after_phase_retrieval_'
>>> tomopy.xtomo_writer(d.data_recon, output_file)
>>> print "Images are succesfully saved at " + output_file + '...'
"""
# Arguments passed by multi-processing wrapper
ind, dshape, inputs = args
# Function inputs
data = mp.tonumpyarray(mp.shared_arr, dshape) # shared-array
pixel_size, dist, energy, alpha, padding = inputs
# Compute the filter.
H, x_shift, y_shift, tmp_proj = paganin_filter(data, pixel_size, dist,
energy, alpha, padding)
num_proj, dx, dy = dshape # dx:slices, dy:pixels
for m in ind:
proj = data[m, :, :]
if padding:
tmp_proj[x_shift:dx + x_shift, y_shift:dy + y_shift] = proj
fft_proj = fftw.fftw2(tmp_proj)
filtered_proj = np.multiply(H, fft_proj)
tmp = np.real(fftw.ifftw2(filtered_proj)) / np.max(H)
proj = tmp[x_shift:dx + x_shift, y_shift:dy + y_shift]
elif not padding:
fft_proj = fftw.fftw2(proj)
filtered_proj = np.multiply(H, fft_proj)
proj = np.real(fftw.ifftw2(filtered_proj)) / np.max(H)
data[m, :, :] = proj
