def evaluate(self):
self.tomo.value = tomopy.remove_outlier(self.tomo.value,
self.dif.value,
size=self.size.value,
axis=self.axis.value,
ncore=self.ncore.value,
out=self.tomo.value)
def filter_one_use_tomopy(img, maxdiff=500, **kwds):
"""does gamma filtering on the given image
This algorithm uses tomopy. maxdiff is the max diff between a pixel
value and median value to be considered normal (not an outlier)
- img: image npy array. must be of integer data type
- kwds: additional kwd args to pass to remove_outliers_bymedian
"""
import tomopy
return tomopy.remove_outlier(img, maxdiff)
def reconstruct(sname, rot_center, ovlpfind, s_start, s_end):
fname = dfolder + sname + '.h5'
print(fname)
start = s_start
end = s_end
chunks = 24
num_sino = (end - start) // chunks
for m in range(chunks):
sino_start = start + num_sino * m
sino_end = start + num_sino * (m + 1)
start_read_time = time.time()
proj, flat, dark, thetat = dxchange.read_aps_2bm(fname,
sino=(sino_start,
sino_end))
print(' done read in %0.1f min' %
((time.time() - start_read_time) / 60))
dark = proj[9001:9002]
flat = proj[0:1]
proj = proj[1:9000]
theta = tomopy.angles(proj.shape[0], 0., 360.)
proj = tomopy.sino_360_to_180(proj, overlap=ovlpfind, rotation='right')
proj = tomopy.remove_outlier(proj, dif=0.4)
proj = tomopy.normalize_bg(proj, air=10)
proj = tomopy.minus_log(proj)
center = rot_center
start_ring_time = time.time()
proj = tomopy.remove_stripe_fw(proj, wname='sym5', sigma=4, pad=False)
proj = tomopy.remove_stripe_sf(proj, size=3)
print(' done pre-process in %0.1f min' %
((time.time() - start_ring_time) / 60))
start_phase_time = time.time()
proj = tomopy.retrieve_phase(proj,
pixel_size=detector_pixel_size_x,
dist=sample_detector_distance,
energy=energy,
alpha=alpha,
pad=True,
ncore=None,
nchunk=None)
print(' done phase retrieval in %0.1f min' %
((time.time() - start_phase_time) / 60))
start_recon_time = time.time()
rec = tomopy.recon(proj,
theta,
center=center,
algorithm='gridrec',
filter_name='ramalk')
tomopy.circ_mask(rec, axis=0, ratio=0.95)
print("Reconstructed", rec.shape)
dxchange.write_tiff_stack(rec,
fname=dfolder + '/' + sname + '/' + sname,
overwrite=True,
start=sino_start)
print(' Chunk reconstruction done in %0.1f min' %
((time.time() - start_recon_time) / 60))
print("Done!")
def recon_hdf5_mpi(src_fanme, dest_folder, sino_range, sino_step, center_vec, shift_grid, dtype='float32',
algorithm='gridrec', tolerance=1, save_sino=False, sino_blur=None, **kwargs):
"""
Reconstruct a single tile, or fused HDF5 created using util/total_fusion. MPI supported.
"""
raise DeprecationWarning
if rank == 0:
if not os.path.exists(dest_folder):
os.mkdir(dest_folder)
sino_ini = int(sino_range[0])
sino_end = int(sino_range[1])
f = h5py.File(src_fanme)
dset = f['exchange/data']
full_shape = dset.shape
theta = tomopy.angles(full_shape[0])
center_vec = np.asarray(center_vec)
sino_ls = np.arange(sino_ini, sino_end, sino_step, dtype='int')
grid_bins = np.ceil(shift_grid[:, 0, 0])
t0 = time.time()
alloc_set = allocate_mpi_subsets(sino_ls.size, size, task_list=sino_ls)
for slice in alloc_set[rank]:
print(' Rank {:d}: reconstructing {:d}'.format(rank, slice))
grid_line = np.digitize(slice, grid_bins)
grid_line = grid_line - 1
center = center_vec[grid_line]
data = dset[:, slice, :]
if sino_blur is not None:
data = gaussian_filter(data, sino_blur)
data = data.reshape([full_shape[0], 1, full_shape[2]])
data[np.isnan(data)] = 0
data = data.astype('float32')
if save_sino:
dxchange.write_tiff(data[:, slice, :], fname=os.path.join(dest_folder, 'sino/recon_{:05d}_{:d}.tiff').format(slice, center))
# data = tomopy.remove_stripe_ti(data)
rec = tomopy.recon(data, theta, center=center, algorithm=algorithm, **kwargs)
# rec = tomopy.remove_ring(rec)
rec = tomopy.remove_outlier(rec, tolerance)
rec = tomopy.circ_mask(rec, axis=0, ratio=0.95)
dxchange.write_tiff(rec, fname='{:s}/recon/recon_{:05d}_{:d}'.format(dest_folder, slice, center), dtype=dtype)
print('Rank {:d} finished in {:.2f} s.'.format(rank, time.time()-t0))
return
def testOutlierRemoval(size):
'''test with random array, loop outside with a 2d cuda kernel'''
# prjsize is z, imsize is x, y.
# size = 13 # window size for the filter
imsizex = 2048 # image size for the input
imsizey = 2048
prjsize = 1
combinedMed = np.zeros(shape=(prjsize, imsizey, imsizex), dtype=np.float32)
resultscuda = np.zeros(shape=(prjsize, imsizey, imsizex), dtype=np.float32)
resultscpu = np.zeros(shape=(prjsize, imsizey, imsizex), dtype=np.float32)
diff = 100
# create combined noise matrix 3D
for step in range(5, 5 + prjsize):
im_noise = np.arange(10, imsizey * imsizex * step + 10,
step).reshape(imsizey, imsizex)
# im_noise = np.random.rand(imsizey, imsizex)
im_noise = im_noise.astype(np.float32)
combinedMed[step - 5] = im_noise
start = timeit.default_timer()
resultscpu = tomopy.remove_outlier_cuda(combinedMed, diff, size=size)
stop = timeit.default_timer()
diff2 = stop - start
print("end cuda outlier removal", diff2)
start = timeit.default_timer()
resultscuda = tomopy.remove_outlier(combinedMed, diff, size=size, ncore=1)
stop = timeit.default_timer()
diff1 = stop - start
print("end cpu outlier removal", diff1)
print("the times gpu over cpu is", diff1 / diff2)
print not np.any(resultscuda - resultscpu)
def testOutlierRemoval(size):
'''test with random array, loop outside with a 2d cuda kernel'''
# prjsize is z, imsize is x, y.
# size = 13 # window size for the filter
imsizex =2048 # image size for the input
imsizey = 2048
prjsize= 1
combinedMed = np.zeros(shape=(prjsize,imsizey,imsizex), dtype=np.float32)
resultscuda = np.zeros(shape=(prjsize,imsizey,imsizex), dtype=np.float32)
resultscpu = np.zeros(shape=(prjsize,imsizey,imsizex), dtype=np.float32)
diff = 100
# create combined noise matrix 3D
for step in range (5, 5 + prjsize):
im_noise = np.arange( 10, imsizey*imsizex*step+10, step ).reshape(imsizey, imsizex)
# im_noise = np.random.rand(imsizey, imsizex)
im_noise = im_noise.astype(np.float32)
combinedMed[step-5]=im_noise
start = timeit.default_timer()
resultscpu= tomopy.remove_outlier_cuda(combinedMed, diff, size=size)
stop = timeit.default_timer()
diff2 = stop - start
print("end cuda outlier removal", diff2)
start = timeit.default_timer()
resultscuda = tomopy.remove_outlier(combinedMed, diff, size=size, ncore = 1)
stop = timeit.default_timer()
diff1 = stop - start
print("end cpu outlier removal", diff1 )
print("the times gpu over cpu is", diff1/diff2)
print not np.any(resultscuda-resultscpu)
def reconstruct(sname, rot_center, ovlpfind, s_start, s_end):
fname = dfolder + sname + '.h5'
print (fname)
start = s_start
end = s_end
chunks = 24
num_sino = (end - start) // chunks
for m in range(chunks):
sino_start = start + num_sino * m
sino_end = start + num_sino * (m + 1)
start_read_time = time.time()
proj, flat, dark, thetat = dxchange.read_aps_2bm(fname, sino=(sino_start, sino_end))
print(' done read in %0.1f min' % ((time.time() - start_read_time)/60))
dark = proj[9001:9002]
flat = proj[0:1]
proj = proj[1:9000]
theta = tomopy.angles(proj.shape[0], 0., 360.)
proj = tomopy.sino_360_to_180(proj, overlap=ovlpfind, rotation='right')
proj = tomopy.remove_outlier(proj, dif=0.4)
proj = tomopy.normalize_bg(proj, air=10)
proj = tomopy.minus_log(proj)
center = rot_center
start_ring_time = time.time()
proj = tomopy.remove_stripe_fw(proj, wname='sym5', sigma=4, pad=False)
proj = tomopy.remove_stripe_sf(proj, size=3)
print(' done pre-process in %0.1f min' % ((time.time() - start_ring_time)/60))
start_phase_time = time.time()
proj = tomopy.retrieve_phase(proj, pixel_size=detector_pixel_size_x, dist=sample_detector_distance, energy=energy, alpha=alpha, pad=True, ncore=None, nchunk=None)
print(' done phase retrieval in %0.1f min' % ((time.time() - start_phase_time)/60))
start_recon_time = time.time()
rec = tomopy.recon(proj, theta, center=center, algorithm='gridrec', filter_name='ramalk')
tomopy.circ_mask(rec, axis=0, ratio=0.95)
print ("Reconstructed", rec.shape)
dxchange.write_tiff_stack(rec, fname = dfolder + '/' + sname + '/' + sname, overwrite=True, start=sino_start)
print(' Chunk reconstruction done in %0.1f min' % ((time.time() - start_recon_time)/60))
print ("Done!")
def recon_block(grid, shift_grid, src_folder, dest_folder, slice_range, sino_step, center_vec, ds_level=0, blend_method='max',
blend_options=None, tolerance=1, sinogram_order=False, algorithm='gridrec', init_recon=None, ncore=None, nchunk=None, dtype='float32',
crop=None, save_sino=False, assert_width=None, sino_blur=None, color_correction=False, flattened_radius=120, normalize=True,
test_mode=False, mode='180', phase_retrieval=None, **kwargs):
"""
Reconstruct dsicrete HDF5 tiles, blending sinograms only.
"""
raw_folder = os.getcwd()
os.chdir(src_folder)
sino_ini = int(slice_range[0])
sino_end = int(slice_range[1])
mod_start_slice = 0
center_vec = np.asarray(center_vec)
center_pos_cache = 0
sino_ls = np.arange(sino_ini, sino_end, sino_step, dtype='int')
pix_shift_grid = np.ceil(shift_grid)
pix_shift_grid[pix_shift_grid < 0] = 0
alloc_set = allocate_mpi_subsets(sino_ls.size, size, task_list=sino_ls)
for i_slice in alloc_set[rank]:
print('############################################')
print('Reconstructing ' + str(i_slice))
# judge from which tile to retrieve sinos
grid_lines = np.zeros(grid.shape[1], dtype=np.int)
slice_in_tile = np.zeros(grid.shape[1], dtype=np.int)
for col in range(grid.shape[1]):
bins = pix_shift_grid[:, col, 0]
grid_lines[col] = int(np.squeeze(np.digitize(i_slice, bins)) - 1)
if grid_lines[col] == -1:
print("WARNING: The specified starting slice number does not allow for full sinogram construction. Trying next slice...")
mod_start_slice = 1
break
else:
mod_start_slice = 0
slice_in_tile[col] = i_slice - bins[grid_lines[col]]
if mod_start_slice == 1:
continue
center_pos = int(np.round(center_vec[grid_lines].mean()))
if center_pos_cache == 0:
center_pos_cache = center_pos
center_diff = center_pos - center_pos_cache
center_pos_0 = center_pos
row_sino, center_pos = prepare_slice(grid, shift_grid, grid_lines, slice_in_tile, ds_level=ds_level,
method=blend_method, blend_options=blend_options, rot_center=center_pos,
assert_width=assert_width, sino_blur=sino_blur, color_correction=color_correction,
normalize=normalize, mode=mode, phase_retrieval=phase_retrieval)
rec0 = recon_slice(row_sino, center_pos, sinogram_order=sinogram_order, algorithm=algorithm,
init_recon=init_recon, ncore=ncore, nchunk=nchunk, **kwargs)
rec = tomopy.remove_ring(np.copy(rec0))
cent = int((rec.shape[1] - 1) / 2)
xx, yy = np.meshgrid(np.arange(rec.shape[2]), np.arange(rec.shape[1]))
mask0 = ((xx - cent) ** 2 + (yy - cent) ** 2 <= flattened_radius ** 2)
mask = np.zeros(rec.shape, dtype='bool')
for i in range(mask.shape[0]):
mask[i, :, :] = mask0
rec[mask] = (rec[mask] + rec0[mask]) / 2
rec = tomopy.remove_outlier(rec, tolerance)
rec = tomopy.circ_mask(rec, axis=0, ratio=0.95)
print('Center: {:d}'.format(center_pos))
rec = np.squeeze(rec)
if center_diff != 0:
rec = np.roll(rec, -center_diff, axis=0)
if not crop is None:
crop = np.asarray(crop)
rec = rec[crop[0, 0]:crop[1, 0], crop[0, 1]:crop[1, 1]]
os.chdir(raw_folder)
if test_mode:
dxchange.write_tiff(rec, fname=os.path.join(dest_folder, 'recon/recon_{:05d}_{:04d}.tiff'.format(i_slice, center_pos)), dtype=dtype)
else:
dxchange.write_tiff(rec, fname=os.path.join(dest_folder, 'recon/recon_{:05d}.tiff'.format(i_slice)), dtype=dtype)
if save_sino:
dxchange.write_tiff(np.squeeze(row_sino), fname=os.path.join(dest_folder, 'sino/sino_{:05d}.tiff'.format(i_slice)), overwrite=True)
os.chdir(src_folder)
os.chdir(raw_folder)
return
datapath = '/home/xf05id1/localdata/CHESSdata/SS_Cl/nf/'
projname = 'SS_Cl_combine_590-664.tif'
samplename = 'SS_Cl'
wf_file_name = '/home/xf05id1/localdata/CHESSdata/darkfield_whitefield/white_field_test_NFA1300_664-590.tif'
#%run /home/xf05id1/src/nsls2-xf-utils/chesstomo_parameters.py
#################
print('loading data')
proj, df, wf = srxfftomo_process.chess_fileio(datapath,
projname,
wf_file=wf_file_name)
print(proj.shape, wf.shape)
proj = tomopy.remove_outlier(proj, 400, size=6)
proj = tomopy.remove_outlier(proj, 100, size=6)
wf = np.expand_dims(
wf, axis=0) #expand the initial 2D white field image into 3D image
wf = tomopy.remove_outlier(wf, 400, size=6)
wf = tomopy.remove_outlier(wf, 100, size=6)
print('correcting background')
proj = srxfftomo_process.srxfftomo_bkg_correction(df, wf, proj)
print('taking negative natural log')
proj = tomopy.minus_log(proj)
print(' handling special values: negatives, Nan, infinite')
proj = tomopy.misc.corr.remove_neg(proj, val=0.001)
proj = tomopy.misc.corr.remove_nan(proj, val=0.001)
def recon_hdf5(src_fanme, dest_folder, sino_range, sino_step, shift_grid, center_vec=None, center_eq=None, dtype='float32',
algorithm='gridrec', tolerance=1, chunk_size=20, save_sino=False, sino_blur=None, flattened_radius=120,
mode='180', test_mode=False, phase_retrieval=None, ring_removal=True, **kwargs):
"""
center_eq: a and b parameters in fitted center position equation center = a*slice + b.
"""
if not os.path.exists(dest_folder):
try:
os.mkdir(dest_folder)
except:
pass
sino_ini = int(sino_range[0])
sino_end = int(sino_range[1])
sino_ls_all = np.arange(sino_ini, sino_end, sino_step, dtype='int')
alloc_set = allocate_mpi_subsets(sino_ls_all.size, size, task_list=sino_ls_all)
sino_ls = alloc_set[rank]
# prepare metadata
f = h5py.File(src_fanme)
dset = f['exchange/data']
full_shape = dset.shape
theta = tomopy.angles(full_shape[0])
if center_eq is not None:
a, b = center_eq
center_ls = sino_ls * a + b
center_ls = np.round(center_ls)
for iblock in range(int(sino_ls.size/chunk_size)+1):
print('Beginning block {:d}.'.format(iblock))
t0 = time.time()
istart = iblock*chunk_size
iend = np.min([(iblock+1)*chunk_size, sino_ls.size])
fstart = sino_ls[istart]
fend = sino_ls[iend]
center = center_ls[istart:iend]
data = dset[:, fstart:fend:sino_step, :]
data[np.isnan(data)] = 0
data = data.astype('float32')
data = tomopy.remove_stripe_ti(data, alpha=4)
if sino_blur is not None:
for i in range(data.shape[1]):
data[:, i, :] = gaussian_filter(data[:, i, :], sino_blur)
rec = tomopy.recon(data, theta, center=center, algorithm=algorithm, **kwargs)
rec = tomopy.remove_ring(rec)
rec = tomopy.remove_outlier(rec, tolerance)
rec = tomopy.circ_mask(rec, axis=0, ratio=0.95)
for i in range(rec.shape[0]):
slice = fstart + i*sino_step
dxchange.write_tiff(rec[i, :, :], fname=os.path.join(dest_folder, 'recon/recon_{:05d}_{:05d}.tiff').format(slice, sino_ini))
if save_sino:
dxchange.write_tiff(data[:, i, :], fname=os.path.join(dest_folder, 'sino/recon_{:05d}_{:d}.tiff').format(slice, int(center[i])))
iblock += 1
print('Block {:d} finished in {:.2f} s.'.format(iblock, time.time()-t0))
else:
# divide chunks
grid_bins = np.append(np.ceil(shift_grid[:, 0, 0]), full_shape[1])
chunks = []
center_ls = []
istart = 0
counter = 0
# irow should be 0 for slice 0
irow = np.searchsorted(grid_bins, sino_ls[0], side='right')-1
for i in range(sino_ls.size):
counter += 1
sino_next = i+1 if i != sino_ls.size-1 else i
if counter >= chunk_size or sino_ls[sino_next] >= grid_bins[irow+1] or sino_next == i:
iend = i+1
chunks.append((istart, iend))
istart = iend
center_ls.append(center_vec[irow])
if sino_ls[sino_next] >= grid_bins[irow+1]:
irow += 1
counter = 0
# reconstruct chunks
iblock = 1
for (istart, iend), center in izip(chunks, center_ls):
print('Beginning block {:d}.'.format(iblock))
t0 = time.time()
fstart = sino_ls[istart]
fend = sino_ls[iend-1]
print('Reading data...')
data = dset[:, fstart:fend+1:sino_step, :]
if mode == '360':
overlap = 2 * (dset.shape[2] - center)
data = tomosaic.morph.sino_360_to_180(data, overlap=overlap, rotation='right')
theta = tomopy.angles(data.shape[0])
data[np.isnan(data)] = 0
data = data.astype('float32')
if sino_blur is not None:
for i in range(data.shape[1]):
data[:, i, :] = gaussian_filter(data[:, i, :], sino_blur)
if ring_removal:
data = tomopy.remove_stripe_ti(data, alpha=4)
if phase_retrieval:
data = tomopy.retrieve_phase(data, kwargs['pixel_size'], kwargs['dist'], kwargs['energy'],
kwargs['alpha'])
rec0 = tomopy.recon(data, theta, center=center, algorithm=algorithm, **kwargs)
rec = tomopy.remove_ring(np.copy(rec0))
cent = int((rec.shape[1]-1) / 2)
xx, yy = np.meshgrid(np.arange(rec.shape[2]), np.arange(rec.shape[1]))
mask0 = ((xx-cent)**2+(yy-cent)**2 <= flattened_radius**2)
mask = np.zeros(rec.shape, dtype='bool')
for i in range(mask.shape[0]):
mask[i, :, :] = mask0
rec[mask] = (rec[mask] + rec0[mask])/2
else:
rec = tomopy.recon(data, theta, center=center, algorithm=algorithm, **kwargs)
rec = tomopy.remove_outlier(rec, tolerance)
rec = tomopy.circ_mask(rec, axis=0, ratio=0.95)
for i in range(rec.shape[0]):
slice = fstart + i*sino_step
if test_mode:
dxchange.write_tiff(rec[i, :, :], fname=os.path.join(dest_folder, 'recon/recon_{:05d}_{:d}.tiff').format(slice, center), dtype=dtype)
else:
dxchange.write_tiff(rec[i, :, :], fname=os.path.join(dest_folder, 'recon/recon_{:05d}.tiff').format(slice), dtype=dtype)
if save_sino:
dxchange.write_tiff(data[:, i, :], fname=os.path.join(dest_folder, 'sino/recon_{:05d}_{:d}.tiff').format(slice, center), dtype=dtype)
print('Block {:d} finished in {:.2f} s.'.format(iblock, time.time()-t0))
iblock += 1
return
def recon_block(grid,
shift_grid,
src_folder,
dest_folder,
slice_range,
sino_step,
center_vec,
ds_level=0,
blend_method='max',
blend_options=None,
tolerance=1,
sinogram_order=False,
algorithm='gridrec',
init_recon=None,
ncore=None,
nchunk=None,
dtype='float32',
crop=None,
save_sino=False,
assert_width=None,
sino_blur=None,
color_correction=False,
flattened_radius=120,
normalize=True,
test_mode=False,
mode='180',
phase_retrieval=None,
data_format='aps_32id',
read_theta=True,
ring_removal=True,
**kwargs):
"""
Reconstruct dsicrete HDF5 tiles, blending sinograms only.
"""
sino_ini = int(slice_range[0])
sino_end = int(slice_range[1])
mod_start_slice = 0
center_vec = np.asarray(center_vec)
center_pos_cache = 0
sino_ls = np.arange(sino_ini, sino_end, sino_step, dtype='int')
alloc_set = allocate_mpi_subsets(sino_ls.size, size, task_list=sino_ls)
for i_slice in alloc_set[rank]:
internal_print('############################################')
internal_print('Reconstructing ' + str(i_slice))
row_sino, center_pos = create_row_sinogram(
grid, shift_grid, src_folder, i_slice, center_vec, ds_level,
blend_method, blend_options, assert_width, sino_blur,
color_correction, normalize, mode, phase_retrieval, data_format)
if row_sino is None:
continue
if read_theta:
_, _, _, theta = read_data_adaptive(os.path.join(
src_folder, grid[0, 0]),
proj=(0, 1),
return_theta=True)
if ring_removal:
rec0 = recon_slice(row_sino,
theta,
center_pos,
sinogram_order=sinogram_order,
algorithm=algorithm,
init_recon=init_recon,
ncore=ncore,
nchunk=nchunk,
**kwargs)
rec = tomopy.remove_ring(np.copy(rec0))
cent = int((rec.shape[1] - 1) / 2)
xx, yy = np.meshgrid(np.arange(rec.shape[2]),
np.arange(rec.shape[1]))
mask0 = ((xx - cent)**2 + (yy - cent)**2 <= flattened_radius**2)
mask = np.zeros(rec.shape, dtype='bool')
for i in range(mask.shape[0]):
mask[i, :, :] = mask0
rec[mask] = (rec[mask] + rec0[mask]) / 2
else:
rec = recon_slice(row_sino,
theta,
center_pos,
sinogram_order=sinogram_order,
algorithm=algorithm,
init_recon=init_recon,
ncore=ncore,
nchunk=nchunk,
**kwargs)
rec = tomopy.remove_outlier(rec, tolerance)
rec = tomopy.circ_mask(rec, axis=0, ratio=0.95)
internal_print('Center: {:d}'.format(center_pos))
rec = np.squeeze(rec)
# correct recon position shifting due to center misalignment
if center_pos_cache == 0:
center_pos_cache = center_pos
center_diff = center_pos - center_pos_cache
if center_diff != 0:
rec = np.roll(rec, -center_diff, axis=0)
if not crop is None:
crop = np.asarray(crop)
rec = rec[crop[0, 0]:crop[1, 0], crop[0, 1]:crop[1, 1]]
if test_mode:
dxchange.write_tiff(rec,
fname=os.path.join(
dest_folder,
'recon/recon_{:05d}_{:04d}.tiff'.format(
i_slice, center_pos)),
dtype=dtype)
else:
dxchange.write_tiff(rec,
fname=os.path.join(
dest_folder,
'recon/recon_{:05d}.tiff'.format(i_slice)),
dtype=dtype)
if save_sino:
dxchange.write_tiff(np.squeeze(row_sino),
fname=os.path.join(
dest_folder,
'sino/sino_{:05d}.tiff'.format(i_slice)),
overwrite=True)
return
def reconstruct(filename,
inputPath="",
outputPath="",
COR=COR,
doOutliers=doOutliers,
outlier_diff=outlier_diff,
outlier_size=outlier_size,
doFWringremoval=doFWringremoval,
ringSigma=ringSigma,
ringLevel=ringLevel,
ringWavelet=ringWavelet,
pad_sino=pad_sino,
doPhaseRetrieval=doPhaseRetrieval,
propagation_dist=propagation_dist,
kev=kev,
alphaReg=alphaReg,
butterworthpars=butterworthpars,
doPolarRing=doPolarRing,
Rarc=Rarc,
Rmaxwidth=Rmaxwidth,
Rtmax=Rtmax,
Rthr=Rthr,
Rtmin=Rtmin,
useAutoCOR=useAutoCOR,
use360to180=use360to180,
num_substacks=num_substacks,
recon_slice=recon_slice):
# Convert filename to list type if only one file name is given
if type(filename) != list:
filename = [filename]
# If useAutoCor == true, a list of COR will be automatically calculated for all files
# If a list of COR is given, only entries with boolean False will use automatic COR calculation
if useAutoCOR == True or (len(COR) != len(filename)):
logging.info('using auto COR for all input files')
COR = [False] * len(filename)
for x in range(len(filename)):
logging.info('opening data set, checking metadata')
fdata, gdata = read_als_832h5_metadata(inputPath[x] + filename[x] +
'.h5')
pxsize = float(
gdata['pxsize']
) / 10.0 # convert from metadata (mm) to this script (cm)
numslices = int(gdata['nslices'])
# recon_slice == True, only center slice will be reconstructed
# if integer is given, a specific
if recon_slice != False:
if (type(recon_slice) == int) and (recon_slice <= numslices):
sinorange[recon_slice - 1, recon_slice]
else:
sinorange = [numslices // 2 - 1, numslices // 2]
else:
sinorange = [0, numslices]
# Calculate number of substacks (chunks)
substacks = num_substacks #(sinorange[1]-sinorange[0]-1)//num_sino_per_substack+1
if (sinorange[1] - sinorange[0]) >= substacks:
num_sino_per_substack = (sinorange[1] -
sinorange[0]) // num_substacks
else:
num_sino_per_substack = 1
firstcor, lastcor = 0, int(gdata['nangles']) - 1
projs, flat, dark, floc = dxchange.read_als_832h5(
inputPath[x] + filename[x] + '.h5', ind_tomo=(firstcor, lastcor))
projs = tomopy.normalize_nf(projs, flat, dark, floc)
autocor = tomopy.find_center_pc(projs[0], projs[1], tol=0.25)
if (type(COR[x]) == bool) or (COR[x] < 0) or (COR[x] == 'auto'):
firstcor, lastcor = 0, int(gdata['nangles']) - 1
projs, flat, dark, floc = dxchange.read_als_832h5(
inputPath[x] + filename[x] + '.h5',
ind_tomo=(firstcor, lastcor))
projs = tomopy.normalize_nf(projs, flat, dark, floc)
cor = tomopy.find_center_pc(projs[0], projs[1], tol=0.25)
else:
cor = COR[x]
logging.info(
'Dataset %s, has %d total slices, reconstructing slices %d through %d in %d substack(s), using COR: %f',
filename[x], int(gdata['nslices']), sinorange[0], sinorange[1] - 1,
substacks, cor)
for y in range(0, substacks):
logging.info('Starting dataset %s (%d of %d), substack %d of %d',
filename[x], x + 1, len(filename), y + 1, substacks)
logging.info('Reading sinograms...')
projs, flat, dark, floc = dxchange.read_als_832h5(
inputPath[x] + filename[x] + '.h5',
sino=(sinorange[0] + y * num_sino_per_substack,
sinorange[0] + (y + 1) * num_sino_per_substack, 1))
logging.info(
'Doing remove outliers, norm (nearest flats), and -log...')
if doOutliers:
projs = tomopy.remove_outlier(projs,
outlier_diff,
size=outlier_size,
axis=0)
flat = tomopy.remove_outlier(flat,
outlier_diff,
size=outlier_size,
axis=0)
tomo = tomopy.normalize_nf(projs, flat, dark, floc)
tomo = tomopy.minus_log(tomo, out=tomo) # in place logarithm
# Use padding to remove halo in reconstruction if present
if pad_sino:
npad = int(
np.ceil(tomo.shape[2] * np.sqrt(2)) - tomo.shape[2]) // 2
tomo = tomopy.pad(tomo, 2, npad=npad, mode='edge')
cor_rec = cor + npad # account for padding
else:
cor_rec = cor
if doFWringremoval:
logging.info('Doing ring (Fourier-wavelet) function...')
tomo = tomopy.remove_stripe_fw(tomo,
sigma=ringSigma,
level=ringLevel,
pad=True,
wname=ringWavelet)
if doPhaseRetrieval:
logging.info('Doing Phase retrieval...')
#tomo = tomopy.retrieve_phase(tomo, pixel_size=pxsize, dist=propagation_dist, energy=kev, alpha=alphaReg, pad=True)
tomo = tomopy.retrieve_phase(tomo,
pixel_size=pxsize,
dist=propagation_dist,
energy=kev,
alpha=alphaReg,
pad=True)
logging.info(
'Doing recon (gridrec) function and scaling/masking, with cor %f...',
cor_rec)
rec = tomopy.recon(tomo,
tomopy.angles(tomo.shape[0], 270, 90),
center=cor_rec,
algorithm='gridrec',
filter_name='butterworth',
filter_par=butterworthpars)
#rec = tomopy.recon(tomo, tomopy.angles(tomo.shape[0], 180+angularrange/2, 180-angularrange/2), center=cor_rec, algorithm='gridrec', filter_name='butterworth', filter_par=butterworthpars)
rec /= pxsize # intensity values in cm^-1
if pad_sino:
rec = tomopy.circ_mask(rec[:, npad:-npad, npad:-npad], 0)
else:
rec = tomopy.circ_mask(rec, 0, ratio=1.0, val=0.0)
if doPolarRing:
logging.info('Doing ring (polar mean filter) function...')
rec = tomopy.remove_ring(rec,
theta_min=Rarc,
rwidth=Rmaxwidth,
thresh_max=Rtmax,
thresh=Rthr,
thresh_min=Rtmin)
logging.info('Writing reconstruction slices to %s', filename[x])
#dxchange.write_tiff_stack(rec, fname=outputPath+'alpha'+str(alphaReg)+'/rec'+filename[x]+'/rec'+filename[x], start=sinorange[0]+y*num_sino_per_substack)
dxchange.write_tiff_stack(rec,
fname=outputPath + 'recon_' +
filename[x] + '/recon_' + filename[x],
start=sinorange[0] +
y * num_sino_per_substack)
logging.info('Reconstruction Complete: ' + filename[x])
def reconstruct(filename,inputPath="", outputPath="", COR=COR, doOutliers=doOutliers, outlier_diff=outlier_diff, outlier_size=outlier_size, doFWringremoval=doFWringremoval, ringSigma=ringSigma,ringLevel=ringLevel, ringWavelet=ringWavelet,pad_sino=pad_sino, doPhaseRetrieval=doPhaseRetrieval, propagation_dist=propagation_dist, kev=kev,alphaReg=alphaReg, butterworthpars=butterworthpars, doPolarRing=doPolarRing,Rarc=Rarc, Rmaxwidth=Rmaxwidth, Rtmax=Rtmax, Rthr=Rthr, Rtmin=Rtmin, useAutoCOR=useAutoCOR, use360to180=use360to180, num_substacks=num_substacks,recon_slice=recon_slice):
# Convert filename to list type if only one file name is given
if type(filename) != list:
filename=[filename]
# If useAutoCor == true, a list of COR will be automatically calculated for all files
# If a list of COR is given, only entries with boolean False will use automatic COR calculation
if useAutoCOR==True or (len(COR) != len(filename)):
logging.info('using auto COR for all input files')
COR = [False]*len(filename)
for x in range(len(filename)):
logging.info('opening data set, checking metadata')
fdata, gdata = read_als_832h5_metadata(inputPath[x]+filename[x]+'.h5')
pxsize = float(gdata['pxsize'])/10.0 # convert from metadata (mm) to this script (cm)
numslices = int(gdata['nslices'])
# recon_slice == True, only center slice will be reconstructed
# if integer is given, a specific
if recon_slice != False:
if (type(recon_slice) == int) and (recon_slice <= numslices):
sinorange [recon_slice-1, recon_slice]
else:
sinorange = [numslices//2-1, numslices//2]
else:
sinorange = [0, numslices]
# Calculate number of substacks (chunks)
substacks = num_substacks #(sinorange[1]-sinorange[0]-1)//num_sino_per_substack+1
if (sinorange[1]-sinorange[0]) >= substacks:
num_sino_per_substack = (sinorange[1]-sinorange[0])//num_substacks
else:
num_sino_per_substack = 1
firstcor, lastcor = 0, int(gdata['nangles'])-1
projs, flat, dark, floc = dxchange.read_als_832h5(inputPath[x]+filename[x]+'.h5', ind_tomo=(firstcor, lastcor))
projs = tomopy.normalize_nf(projs, flat, dark, floc)
autocor = tomopy.find_center_pc(projs[0], projs[1], tol=0.25)
if (type(COR[x]) == bool) or (COR[x]<0) or (COR[x]=='auto'):
firstcor, lastcor = 0, int(gdata['nangles'])-1
projs, flat, dark, floc = dxchange.read_als_832h5(inputPath[x]+filename[x]+'.h5', ind_tomo=(firstcor, lastcor))
projs = tomopy.normalize_nf(projs, flat, dark, floc)
cor = tomopy.find_center_pc(projs[0], projs[1], tol=0.25)
else:
cor = COR[x]
logging.info('Dataset %s, has %d total slices, reconstructing slices %d through %d in %d substack(s), using COR: %f',filename[x], int(gdata['nslices']), sinorange[0], sinorange[1]-1, substacks, cor)
for y in range(0, substacks):
logging.info('Starting dataset %s (%d of %d), substack %d of %d',filename[x], x+1, len(filename), y+1, substacks)
logging.info('Reading sinograms...')
projs, flat, dark, floc = dxchange.read_als_832h5(inputPath[x]+filename[x]+'.h5', sino=(sinorange[0]+y*num_sino_per_substack, sinorange[0]+(y+1)*num_sino_per_substack, 1))
logging.info('Doing remove outliers, norm (nearest flats), and -log...')
if doOutliers:
projs = tomopy.remove_outlier(projs, outlier_diff, size=outlier_size, axis=0)
flat = tomopy.remove_outlier(flat, outlier_diff, size=outlier_size, axis=0)
tomo = tomopy.normalize_nf(projs, flat, dark, floc)
tomo = tomopy.minus_log(tomo, out=tomo) # in place logarithm
# Use padding to remove halo in reconstruction if present
if pad_sino:
npad = int(np.ceil(tomo.shape[2] * np.sqrt(2)) - tomo.shape[2])//2
tomo = tomopy.pad(tomo, 2, npad=npad, mode='edge')
cor_rec = cor + npad # account for padding
else:
cor_rec = cor
if doFWringremoval:
logging.info('Doing ring (Fourier-wavelet) function...')
tomo = tomopy.remove_stripe_fw(tomo, sigma=ringSigma, level=ringLevel, pad=True, wname=ringWavelet)
if doPhaseRetrieval:
logging.info('Doing Phase retrieval...')
#tomo = tomopy.retrieve_phase(tomo, pixel_size=pxsize, dist=propagation_dist, energy=kev, alpha=alphaReg, pad=True)
tomo = tomopy.retrieve_phase(tomo, pixel_size=pxsize, dist=propagation_dist, energy=kev, alpha=alphaReg, pad=True)
logging.info('Doing recon (gridrec) function and scaling/masking, with cor %f...',cor_rec)
rec = tomopy.recon(tomo, tomopy.angles(tomo.shape[0], 270, 90), center=cor_rec, algorithm='gridrec', filter_name='butterworth', filter_par=butterworthpars)
#rec = tomopy.recon(tomo, tomopy.angles(tomo.shape[0], 180+angularrange/2, 180-angularrange/2), center=cor_rec, algorithm='gridrec', filter_name='butterworth', filter_par=butterworthpars)
rec /= pxsize # intensity values in cm^-1
if pad_sino:
rec = tomopy.circ_mask(rec[:, npad:-npad, npad:-npad], 0)
else:
rec = tomopy.circ_mask(rec, 0, ratio=1.0, val=0.0)
if doPolarRing:
logging.info('Doing ring (polar mean filter) function...')
rec = tomopy.remove_ring(rec, theta_min=Rarc, rwidth=Rmaxwidth, thresh_max=Rtmax, thresh=Rthr, thresh_min=Rtmin)
logging.info('Writing reconstruction slices to %s', filename[x])
#dxchange.write_tiff_stack(rec, fname=outputPath+'alpha'+str(alphaReg)+'/rec'+filename[x]+'/rec'+filename[x], start=sinorange[0]+y*num_sino_per_substack)
dxchange.write_tiff_stack(rec, fname=outputPath + 'recon_'+filename[x]+'/recon_'+filename[x], start=sinorange[0]+y*num_sino_per_substack)
logging.info('Reconstruction Complete: '+ filename[x])
def do_work(out_path, scan_id, center_offset, mpi_projs, flats, darks,
mpi_thetas):
logger.info("Flat Field Correction")
mpi_projs = normalize_mpi(mpi_projs, flats, darks)
#utils_mpi.write_stack_mpi(out_path/"flat", mpi_projs)
del flats
del darks
logger.info("Outlier Removal")
# TODO: base parameters on clean simulation data! - might need fill
projs = mpi_projs.scatter(0)
# TODO: put back in, I think it is causing issues right now...
tomopy.remove_outlier(projs, 0.1, 5, ncore=ncore, out=projs)
#tomopy.remove_outlier_cuda(projs, 0.1, 5, ncore, out=projs)
np.clip(projs, 1E-6, 1 - 1E-6, projs)
# TODO: distortion correction factor?
# TODO: ring removal?
# # flat field change correction
# remove_low_frequency_mpi(mpi_projs)
# utils_mpi.write_stack_mpi(out_path/"low_frequency_removed", mpi_projs)
# bulk Si intensity correction
# removes constant absorption contribution from bulk Si, and mounting material
# TODO: base parameters on clean simulation data! - will need fill
# TODO: alternatively, refine result after good recon - with theta offset
target_transmission = 0.80
logger.info(f"Setting target transmission to {target_transmission}")
set_target_transmission_mpi(mpi_projs, mpi_thetas, target_transmission)
projs = mpi_projs.scatter(0)
np.clip(projs, 1E-6, 1 - 1E-6, projs)
utils_mpi.write_stack_mpi(out_path / "constant_transmission", mpi_projs)
# center finding - manual for now?
if center_offset is None:
logger.info("Finding center")
# algorithm = "SART"
# pixel_size = 2 * 0.000016 #16nm bin 1
# options = {"PixelWidth": pixel_size,
# "PixelHeight": pixel_size,
# "windowFOV": False,
# "archDir": out_path,
# "_mpi_rank": mpi_rank,
# }
# alg_params = {"N_iter": 1,
# "N_subsets": 20,
# "nonnegativityConstraint": True,
# "useFBPasSeedImage": False,
# # "Preconditioner": "RAMP",
# # "beta": 2e-7,
# # "p": 1,
# # "delta": 1/20, # delta sets edge strength (difference between regions divide by ten)
# # "inverseVarianceExponent": 1.0, # set to 1 to include noise model
# # "other": 3, #convergence of low frequencies
# }
# # load data into LTT, then find the center before recon
# ltt_tomopy.initialize_recon(sinos, thetas, xcenter, True, algorithm, options, ncore=ncore)
# center = align_tomo.find_center_ltt(lambda c: ltt_tomopy.preview(center=c, algorithm=algorithm, sinogram_order=True, close=False, options=options, alg_params=alg_params, ncore=ncore), xcenter, 0.1, ratio=0.8)
# ltt_tomopy.recon_close()
logger.info("Padding sinos for center finding")
mpi_sinos = utils_mpi.create_sinos_mpi(mpi_projs, ncore)
sinos = mpi_sinos.scatter(0)
sinos = pad_sinos(sinos)
mpi_sinos = MpiArray(sinos)
gthetas = mpi_thetas.allgather()
xcenter = sinos.shape[2] // 2
cen_range = (xcenter - 20, xcenter + 20, 0.5)
if mpi_rank == mpi_size // 2:
tomopy.write_center(sinos,
gthetas,
out_path / ("center"),
cen_range,
sinogram_order=True)
del mpi_sinos, sinos
import sys
comm.Barrier()
sys.exit()
# center_offset = mpi_projs.shape[2]//2-center
# mpi_projs.comm.Barrier() #for printing
# print(f"{mpi_projs.mpi_rank}: center={center} offset={center_offset}")
#center = xcenter + center_offset
# Shift correction?
# use MPI and binning for speed
# use LTT for all recon
# define recon extent with extra X and less Z
# Quick recon - LTT with SART or other? (can't use FBP)
algorithm = "gridrec"
options = {
"filter_name": "parzen",
}
logger.info(f"Finding layer alignment within volume using {algorithm}")
mpi_rec = tomopy_recon_mpi(mpi_projs,
mpi_thetas,
center_offset,
algorithm,
ncore=ncore,
**options)
utils_mpi.write_stack_mpi(out_path / ("quick_" + algorithm), mpi_rec)
theta_deg, start_z1, end_z1 = align_layers.find_angle_mpi(mpi_rec, 0)
logger.info(f"Theta offset angle {theta_deg:0.2} deg")
# find phi angle (axis 2)
phi_deg, start_z2, end_z2 = align_layers.find_angle_mpi(mpi_rec, 2)
logger.info(f"Phi offset angle {phi_deg:0.2} deg")
start_z = min(start_z1, start_z2)
end_z = max(end_z1, end_z2)
# add buffer for start and end
start_z = max(start_z - 20, 0)
end_z = min(end_z + 20, mpi_rec.shape[1] - 1)
#FIXME: override start and end
start_z = 0
end_z = mpi_rec.shape[1] - 1
logger.info(f"Layer extent: {start_z} - {end_z}")
# change theta with correction for next reconstruction
thetas = mpi_thetas.scatter(0)
thetas += np.deg2rad(theta_deg)
# modify projections to remove phi angle
# TODO: combine with stage shift code
projs = mpi_projs.scatter(0)
align_layers.apply_phi_correction(projs, thetas, phi_deg, projs)
# Quick aligned recon
algorithm = "gridrec"
options = {
"filter_name": "parzen",
}
logger.info("Quick Tomopy Recon")
mpi_rec = tomopy_recon_mpi(mpi_projs,
mpi_thetas,
center_offset,
algorithm,
ncore=ncore,
**options)
rec = mpi_rec.scatter(0)
rec = rec[:, start_z:end_z, :]
mpi_rec = MpiArray(rec)
utils_mpi.write_stack_mpi(out_path / algorithm, mpi_rec)
del mpi_rec, rec
# Aligned recon
# iterative recon with extra recon space in X and a restricted Z axis
algorithm = "SART" #"RDLS"#"DFM"#"ASD-POCS"#"SART"#"FBP"
logger.info(f"Reconstructing aligned layers using {algorithm}")
mpi_sinos = utils_mpi.create_sinos_mpi(mpi_projs, ncore)
#utils_mpi.write_stack_mpi(out_path/"sinos", mpi_sinos)
sinos = mpi_sinos.scatter(0)
# add padding to recon - fixes cupping effect
xrecpadding = sinos.shape[2] // 2
pixel_size = 2 * 0.000016 # 16nm bin 1
options = {
"PixelWidth": pixel_size,
"PixelHeight": pixel_size,
"ryoffset": start_z,
"ryelements": end_z - start_z,
"windowFOV": False,
"rxelements": sinos.shape[2] + 2 * xrecpadding,
"rxoffset": -xrecpadding,
"_mpi_rank": mpi_rank,
}
alg_params = {
"N_iter": 50,
"nonnegativityConstraint": False,
"useFBPasSeedImage": False,
#"Preconditioner": "RAMP",
#"descentType": "CG",#"GD",
#"beta": 2e-7,
#"p": 1,
#"delta": 20/20, # delta sets edge strength (difference between regions divide by ten)
#"inverseVarianceExponent": 1.0, # set to 1 to include noise model
#"other": 3, #convergence of low frequencies
}
#TODO: add support to add overlap in future with updates between iterations (see xray_trust6.py)
gthetas = mpi_thetas.allgather() #global thetas
center = sinos.shape[2] // 2 + center_offset
if gthetas[1] < gthetas[0]:
# decreasing angle, LTT doesn't support, switch data around
# TODO: load in reversed order?
gthetas = gthetas[::-1]
sinos[:] = sinos[:, ::-1, :]
rec = ltt_tomopy.recon(sinos,
gthetas,
center,
True,
algorithm,
alg_params,
options,
ncore=ncore)
rec = rec[:, :, xrecpadding:xrecpadding + sinos.shape[2]]
mpi_rec = MpiArray(rec, distribution=mpi_sinos.distribution)
utils_mpi.write_stack_mpi(out_path / algorithm, mpi_rec)
# Neural network processing?
# Extract layers
# Use template if available
logger.info("Extracting Layers")
mpi_layers = align_layers.extract_layers_mpi(mpi_rec)
align_layers.write_layers_to_file_mpi(mpi_layers, "layers")
logger.info(f"Finished {scan_id}")
def recon(
filename,
inputPath='./',
outputPath=None,
outputFilename=None,
doOutliers1D=False, # outlier removal in 1d (along sinogram columns)
outlier_diff1D=750, # difference between good data and outlier data (outlier removal)
outlier_size1D=3, # radius around each pixel to look for outliers (outlier removal)
doOutliers2D=False, # outlier removal, standard 2d on each projection
outlier_diff2D=750, # difference between good data and outlier data (outlier removal)
outlier_size2D=3, # radius around each pixel to look for outliers (outlier removal)
doFWringremoval=True, # Fourier-wavelet ring removal
doTIringremoval=False, # Titarenko ring removal
doSFringremoval=False, # Smoothing filter ring removal
ringSigma=3, # damping parameter in Fourier space (Fourier-wavelet ring removal)
ringLevel=8, # number of wavelet transform levels (Fourier-wavelet ring removal)
ringWavelet='db5', # type of wavelet filter (Fourier-wavelet ring removal)
ringNBlock=0, # used in Titarenko ring removal (doTIringremoval)
ringAlpha=1.5, # used in Titarenko ring removal (doTIringremoval)
ringSize=5, # used in smoothing filter ring removal (doSFringremoval)
doPhaseRetrieval=False, # phase retrieval
alphaReg=0.0002, # smaller = smoother (used for phase retrieval)
propagation_dist=75, # sample-to-scintillator distance (phase retrieval)
kev=24, # energy level (phase retrieval)
butterworth_cutoff=0.25, #0.1 would be very smooth, 0.4 would be very grainy (reconstruction)
butterworth_order=2, # for reconstruction
doPolarRing=False, # ring removal
Rarc=30, # min angle needed to be considered ring artifact (ring removal)
Rmaxwidth=100, # max width of rings to be filtered (ring removal)
Rtmax=3000.0, # max portion of image to filter (ring removal)
Rthr=3000.0, # max value of offset due to ring artifact (ring removal)
Rtmin=-3000.0, # min value of image to filter (ring removal)
cor=None, # center of rotation (float). If not used then cor will be detected automatically
corFunction='pc', # center of rotation function to use - can be 'pc', 'vo', or 'nm'
voInd=None, # index of slice to use for cor search (vo)
voSMin=-40, # min radius for searching in sinogram (vo)
voSMax=40, # max radius for searching in sinogram (vo)
voSRad=10, # search radius (vo)
voStep=0.5, # search step (vo)
voRatio=2.0, # ratio of field-of-view and object size (vo)
voDrop=20, # drop lines around vertical center of mask (vo)
nmInd=None, # index of slice to use for cor search (nm)
nmInit=None, # initial guess for center (nm)
nmTol=0.5, # desired sub-pixel accuracy (nm)
nmMask=True, # if True, limits analysis to circular region (nm)
nmRatio=1.0, # ratio of radius of circular mask to edge of reconstructed image (nm)
nmSinoOrder=False, # if True, analyzes in sinogram space. If False, analyzes in radiograph space
use360to180=False, # use 360 to 180 conversion
doBilateralFilter=False, # if True, uses bilateral filter on image just before write step # NOTE: image will be converted to 8bit if it is not already
bilateral_srad=3, # spatial radius for bilateral filter (image will be converted to 8bit if not already)
bilateral_rrad=30, # range radius for bilateral filter (image will be converted to 8bit if not already)
castTo8bit=False, # convert data to 8bit before writing
cast8bit_min=-10, # min value if converting to 8bit
cast8bit_max=30, # max value if converting to 8bit
useNormalize_nf=False, # normalize based on background intensity (nf)
chunk_proj=100, # chunk size in projection direction
chunk_sino=100, # chunk size in sinogram direction
npad=None, # amount to pad data before reconstruction
projused=None, #should be slicing in projection dimension (start,end,step)
sinoused=None, #should be sliceing in sinogram dimension (start,end,step). If first value is negative, it takes the number of slices from the second value in the middle of the stack.
correcttilt=0, #tilt dataset
tiltcenter_slice=None, # tilt center (x direction)
tiltcenter_det=None, # tilt center (y direction)
angle_offset=0, #this is the angle offset from our default (270) so that tomopy yields output in the same orientation as previous software (Octopus)
anglelist=None, #if not set, will assume evenly spaced angles which will be calculated by the angular range and number of angles found in the file. if set to -1, will read individual angles from each image. alternatively, a list of angles can be passed.
doBeamHardening=False, #turn on beam hardening correction, based on "Correction for beam hardening in computed tomography", Gabor Herman, 1979 Phys. Med. Biol. 24 81
BeamHardeningCoefficients=None, #6 values, tomo = a0 + a1*tomo + a2*tomo^2 + a3*tomo^3 + a4*tomo^4 + a5*tomo^5
projIgnoreList=None, #projections to be ignored in the reconstruction (for simplicity in the code, they will not be removed and will be processed as all other projections but will be set to zero absorption right before reconstruction.
*args,
**kwargs):
start_time = time.time()
print("Start {} at:".format(filename) +
time.strftime("%a, %d %b %Y %H:%M:%S +0000", time.localtime()))
outputPath = inputPath if outputPath is None else outputPath
outputFilename = filename if outputFilename is None else outputFilename
tempfilenames = [outputPath + 'tmp0.h5', outputPath + 'tmp1.h5']
filenametowrite = outputPath + '/rec' + filename.strip(
".h5") + '/' + outputFilename
#filenametowrite = outputPath+'/rec'+filename+'/'+outputFilename
print("cleaning up previous temp files", end="")
for tmpfile in tempfilenames:
try:
os.remove(tmpfile)
except OSError:
pass
print(", reading metadata")
datafile = h5py.File(inputPath + filename, 'r')
gdata = dict(dxchange.reader._find_dataset_group(datafile).attrs)
pxsize = float(gdata['pxsize']) / 10 # /10 to convert unites from mm to cm
numslices = int(gdata['nslices'])
numangles = int(gdata['nangles'])
angularrange = float(gdata['arange'])
numrays = int(gdata['nrays'])
npad = int(np.ceil(numrays * np.sqrt(2)) -
numrays) // 2 if npad is None else npad
projused = (0, numangles - 1, 1) if projused is None else projused
# ndark = int(gdata['num_dark_fields'])
# ind_dark = list(range(0, ndark))
# group_dark = [numangles - 1]
inter_bright = int(gdata['i0cycle'])
nflat = int(gdata['num_bright_field'])
ind_flat = list(range(0, nflat))
if inter_bright > 0:
group_flat = list(range(0, numangles, inter_bright))
if group_flat[-1] != numangles - 1:
group_flat.append(numangles - 1)
elif inter_bright == 0:
group_flat = [0, numangles - 1]
else:
group_flat = None
ind_tomo = list(range(0, numangles))
floc_independent = dxchange.reader._map_loc(ind_tomo, group_flat)
#figure out the angle list (a list of angles, one per projection image)
dtemp = datafile[list(datafile.keys())[0]]
fltemp = list(dtemp.keys())
firstangle = float(dtemp[fltemp[0]].attrs.get('rot_angle', 0))
if anglelist is None:
#the offset angle should offset from the angle of the first image, which is usually 0, but in the case of timbir data may not be.
#we add the 270 to be inte same orientation as previous software used at bl832
angle_offset = 270 + angle_offset - firstangle
anglelist = tomopy.angles(numangles, angle_offset,
angle_offset - angularrange)
elif anglelist == -1:
anglelist = np.zeros(shape=numangles)
for icount in range(0, numangles):
anglelist[icount] = np.pi / 180 * (270 + angle_offset - float(
dtemp[fltemp[icount]].attrs['rot_angle']))
#if projused is different than default, need to chnage numangles and angularrange
#can't do useNormalize_nf and doOutliers2D at the same time, or doOutliers2D and doOutliers1D at the same time, b/c of the way we chunk, for now just disable that
if useNormalize_nf == True and doOutliers2D == True:
useNormalize_nf = False
print(
"we cannot currently do useNormalize_nf and doOutliers2D at the same time, turning off useNormalize_nf"
)
if doOutliers2D == True and doOutliers1D == True:
doOutliers1D = False
print(
"we cannot currently do doOutliers1D and doOutliers2D at the same time, turning off doOutliers1D"
)
#figure out how user can pass to do central x number of slices, or set of slices dispersed throughout (without knowing a priori the value of numslices)
if sinoused is None:
sinoused = (0, numslices, 1)
elif sinoused[0] < 0:
sinoused = (
int(np.floor(numslices / 2.0) - np.ceil(sinoused[1] / 2.0)),
int(np.floor(numslices / 2.0) + np.floor(sinoused[1] / 2.0)), 1)
num_proj_per_chunk = np.minimum(chunk_proj, projused[1] - projused[0])
numprojchunks = (projused[1] - projused[0] - 1) // num_proj_per_chunk + 1
num_sino_per_chunk = np.minimum(chunk_sino, sinoused[1] - sinoused[0])
numsinochunks = (sinoused[1] - sinoused[0] - 1) // num_sino_per_chunk + 1
numprojused = (projused[1] - projused[0]) // projused[2]
numsinoused = (sinoused[1] - sinoused[0]) // sinoused[2]
BeamHardeningCoefficients = (
0, 1, 0, 0, 0,
.1) if BeamHardeningCoefficients is None else BeamHardeningCoefficients
if cor is None:
print("Detecting center of rotation", end="")
if angularrange > 300:
lastcor = int(np.floor(numangles / 2) - 1)
else:
lastcor = numangles - 1
#I don't want to see the warnings about the reader using a deprecated variable in dxchange
with warnings.catch_warnings():
warnings.simplefilter("ignore")
tomo, flat, dark, floc = dxchange.read_als_832h5(
inputPath + filename, ind_tomo=(0, lastcor))
tomo = tomo.astype(np.float32)
if useNormalize_nf:
tomopy.normalize_nf(tomo, flat, dark, floc, out=tomo)
else:
tomopy.normalize(tomo, flat, dark, out=tomo)
if corFunction == 'vo':
# same reason for catching warnings as above
with warnings.catch_warnings():
warnings.simplefilter("ignore")
cor = tomopy.find_center_vo(tomo,
ind=voInd,
smin=voSMin,
smax=voSMax,
srad=voSRad,
step=voStep,
ratio=voRatio,
drop=voDrop)
elif corFunction == 'nm':
cor = tomopy.find_center(
tomo,
tomopy.angles(numangles, angle_offset,
angle_offset - angularrange),
ind=nmInd,
init=nmInit,
tol=nmTol,
mask=nmMask,
ratio=nmRatio,
sinogram_order=nmSinoOrder)
elif corFunction == 'pc':
cor = tomopy.find_center_pc(tomo[0], tomo[1], tol=0.25)
else:
raise ValueError("\'corFunction\' must be one of: [ pc, vo, nm ].")
print(", {}".format(cor))
else:
print("using user input center of {}".format(cor))
function_list = []
if doOutliers1D:
function_list.append('remove_outlier1d')
if doOutliers2D:
function_list.append('remove_outlier2d')
if useNormalize_nf:
function_list.append('normalize_nf')
else:
function_list.append('normalize')
function_list.append('minus_log')
if doBeamHardening:
function_list.append('beam_hardening')
if doFWringremoval:
function_list.append('remove_stripe_fw')
if doTIringremoval:
function_list.append('remove_stripe_ti')
if doSFringremoval:
function_list.append('remove_stripe_sf')
if correcttilt:
function_list.append('correcttilt')
if use360to180:
function_list.append('do_360_to_180')
if doPhaseRetrieval:
function_list.append('phase_retrieval')
function_list.append('recon_mask')
if doPolarRing:
function_list.append('polar_ring')
if castTo8bit:
function_list.append('castTo8bit')
if doBilateralFilter:
function_list.append('bilateral_filter')
function_list.append('write_output')
# Figure out first direction to slice
for func in function_list:
if slice_dir[func] != 'both':
axis = slice_dir[func]
break
done = False
curfunc = 0
curtemp = 0
while True: # Loop over reading data in certain chunking direction
if axis == 'proj':
niter = numprojchunks
else:
niter = numsinochunks
for y in range(niter): # Loop over chunks
print("{} chunk {} of {}".format(axis, y + 1, niter))
if curfunc == 0:
with warnings.catch_warnings():
warnings.simplefilter("ignore")
if axis == 'proj':
tomo, flat, dark, floc = dxchange.read_als_832h5(
inputPath + filename,
ind_tomo=range(
y * num_proj_per_chunk + projused[0],
np.minimum(
(y + 1) * num_proj_per_chunk + projused[0],
numangles)),
sino=(sinoused[0], sinoused[1], sinoused[2]))
else:
tomo, flat, dark, floc = dxchange.read_als_832h5(
inputPath + filename,
ind_tomo=range(projused[0], projused[1],
projused[2]),
sino=(y * num_sino_per_chunk + sinoused[0],
np.minimum((y + 1) * num_sino_per_chunk +
sinoused[0], numslices), 1))
else:
if axis == 'proj':
start, end = y * num_proj_per_chunk, np.minimum(
(y + 1) * num_proj_per_chunk, numprojused)
tomo = dxchange.reader.read_hdf5(
tempfilenames[curtemp],
'/tmp/tmp',
slc=((start, end, 1), (0, numslices, 1),
(0, numrays, 1))) #read in intermediate file
else:
start, end = y * num_sino_per_chunk, np.minimum(
(y + 1) * num_sino_per_chunk, numsinoused)
tomo = dxchange.reader.read_hdf5(tempfilenames[curtemp],
'/tmp/tmp',
slc=((0, numangles,
1), (start, end, 1),
(0, numrays, 1)))
dofunc = curfunc
keepvalues = None
while True: # Loop over operations to do in current chunking direction
func_name = function_list[dofunc]
newaxis = slice_dir[func_name]
if newaxis != 'both' and newaxis != axis:
# We have to switch axis, so flush to disk
if y == 0:
try:
os.remove(tempfilenames[1 - curtemp])
except OSError:
pass
appendaxis = 1 if axis == 'sino' else 0
dxchange.writer.write_hdf5(
tomo,
fname=tempfilenames[1 - curtemp],
gname='tmp',
dname='tmp',
overwrite=False,
appendaxis=appendaxis) #writing intermediate file...
break
print(func_name, end=" ")
curtime = time.time()
if func_name == 'remove_outlier1d':
tomo = tomo.astype(np.float32, copy=False)
remove_outlier1d(tomo,
outlier_diff1D,
size=outlier_size1D,
out=tomo)
if func_name == 'remove_outlier2d':
tomo = tomo.astype(np.float32, copy=False)
tomopy.remove_outlier(tomo,
outlier_diff2D,
size=outlier_size2D,
axis=0,
out=tomo)
elif func_name == 'normalize_nf':
tomo = tomo.astype(np.float32, copy=False)
tomopy.normalize_nf(
tomo, flat, dark, floc_independent, out=tomo
) #use floc_independent b/c when you read file in proj chunks, you don't get the correct floc returned right now to use here.
elif func_name == 'normalize':
tomo = tomo.astype(np.float32, copy=False)
tomopy.normalize(tomo, flat, dark, out=tomo)
elif func_name == 'minus_log':
mx = np.float32(0.00000000000000000001)
ne.evaluate('where(tomo>mx, tomo, mx)', out=tomo)
tomopy.minus_log(tomo, out=tomo)
elif func_name == 'beam_hardening':
loc_dict = {
'a{}'.format(i): np.float32(val)
for i, val in enumerate(BeamHardeningCoefficients)
}
tomo = ne.evaluate(
'a0 + a1*tomo + a2*tomo**2 + a3*tomo**3 + a4*tomo**4 + a5*tomo**5',
local_dict=loc_dict,
out=tomo)
elif func_name == 'remove_stripe_fw':
tomo = tomopy.remove_stripe_fw(tomo,
sigma=ringSigma,
level=ringLevel,
pad=True,
wname=ringWavelet)
elif func_name == 'remove_stripe_ti':
tomo = tomopy.remove_stripe_ti(tomo,
nblock=ringNBlock,
alpha=ringAlpha)
elif func_name == 'remove_stripe_sf':
tomo = tomopy.remove_stripe_sf(tomo, size=ringSize)
elif func_name == 'correcttilt':
if tiltcenter_slice is None:
tiltcenter_slice = numslices / 2.
if tiltcenter_det is None:
tiltcenter_det = tomo.shape[2] / 2
new_center = tiltcenter_slice - 0.5 - sinoused[0]
center_det = tiltcenter_det - 0.5
#add padding of 10 pixels, to be unpadded right after tilt correction. This makes the tilted image not have zeros at certain edges, which matters in cases where sample is bigger than the field of view. For the small amounts we are generally tilting the images, 10 pixels is sufficient.
# tomo = tomopy.pad(tomo, 2, npad=10, mode='edge')
# center_det = center_det + 10
cntr = (center_det, new_center)
for b in range(tomo.shape[0]):
tomo[b] = st.rotate(
tomo[b],
correcttilt,
center=cntr,
preserve_range=True,
order=1,
mode='edge',
clip=True
) #center=None means image is rotated around its center; order=1 is default, order of spline interpolation
# tomo = tomo[:, :, 10:-10]
elif func_name == 'do_360_to_180':
# Keep values around for processing the next chunk in the list
keepvalues = [
angularrange, numangles, projused, num_proj_per_chunk,
numprojchunks, numprojused, numrays, anglelist
]
#why -.5 on one and not on the other?
if tomo.shape[0] % 2 > 0:
tomo = sino_360_to_180(
tomo[0:-1, :, :],
overlap=int(
np.round((tomo.shape[2] - cor - .5)) * 2),
rotation='right')
angularrange = angularrange / 2 - angularrange / (
tomo.shape[0] - 1)
else:
tomo = sino_360_to_180(
tomo[:, :, :],
overlap=int(np.round((tomo.shape[2] - cor)) * 2),
rotation='right')
angularrange = angularrange / 2
numangles = int(numangles / 2)
projused = (0, numangles - 1, 1)
num_proj_per_chunk = np.minimum(chunk_proj,
projused[1] - projused[0])
numprojchunks = (projused[1] - projused[0] -
1) // num_proj_per_chunk + 1
numprojused = (projused[1] - projused[0]) // projused[2]
numrays = tomo.shape[2]
anglelist = anglelist[:numangles]
elif func_name == 'phase_retrieval':
tomo = tomopy.retrieve_phase(tomo,
pixel_size=pxsize,
dist=propagation_dist,
energy=kev,
alpha=alphaReg,
pad=True)
elif func_name == 'recon_mask':
tomo = tomopy.pad(tomo, 2, npad=npad, mode='edge')
if projIgnoreList is not None:
for badproj in projIgnoreList:
tomo[badproj] = 0
rec = tomopy.recon(
tomo,
anglelist,
center=cor + npad,
algorithm='gridrec',
filter_name='butterworth',
filter_par=[butterworth_cutoff, butterworth_order])
rec = rec[:, npad:-npad, npad:-npad]
rec /= pxsize # convert reconstructed voxel values from 1/pixel to 1/cm
rec = tomopy.circ_mask(rec, 0)
elif func_name == 'polar_ring':
rec = np.ascontiguousarray(rec, dtype=np.float32)
rec = tomopy.remove_ring(rec,
theta_min=Rarc,
rwidth=Rmaxwidth,
thresh_max=Rtmax,
thresh=Rthr,
thresh_min=Rtmin,
out=rec)
elif func_name == 'castTo8bit':
rec = convert8bit(rec, cast8bit_min, cast8bit_max)
elif func_name == 'bilateral_filter':
rec = pyF3D.run_BilateralFilter(
rec,
spatialRadius=bilateral_srad,
rangeRadius=bilateral_rrad)
elif func_name == 'write_output':
dxchange.write_tiff_stack(rec,
fname=filenametowrite,
start=y * num_sino_per_chunk +
sinoused[0])
print('(took {:.2f} seconds)'.format(time.time() - curtime))
dofunc += 1
if dofunc == len(function_list):
break
if y < niter - 1 and keepvalues: # Reset original values for next chunk
angularrange, numangles, projused, num_proj_per_chunk, numprojchunks, numprojused, numrays, anglelist = keepvalues
curtemp = 1 - curtemp
curfunc = dofunc
if curfunc == len(function_list):
break
axis = slice_dir[function_list[curfunc]]
print("cleaning up temp files")
for tmpfile in tempfilenames:
try:
os.remove(tmpfile)
except OSError:
pass
print("End Time: " +
time.strftime("%a, %d %b %Y %H:%M:%S +0000", time.localtime()))
print('It took {:.3f} s to process {}'.format(time.time() - start_time,
inputPath + filename))
last_sino_slice = sinorange[0]+(y+1)*num_sino_per_chunk
else:
first_sino_slice = np.maximum(sinorange[0]+y*num_sino_per_chunk-overlap_chunk, sinorange[0])
last_sino_slice = np.minimum(sinorange[0]+(y+1)*num_sino_per_chunk+overlap_chunk, sinorange[1])
projs, flat, dark, flocNotUsedSeeOtherFLOCvar = dxchange.read_als_832h5(idirectory+iname[x]+'.h5', sino=(first_sino_slice, last_sino_slice, 1))
if specialBakCrop[x]>0:
num_per_flat = flat.shape[0]//original_floc_length
#logging.info('length flats before: %d',flat.shape[0])
flat = flat[0:-specialBakCrop[x]*num_per_flat]
#logging.info('length flats after: %d',flat.shape[0])
if doOutliers:
projs = tomopy.remove_outlier(projs, outlier_diff, size=outlier_size, axis=0)
flat = tomopy.remove_outlier(flat, outlier_diff, size=outlier_size, axis=0)
if useNormalize_nf:
logging.info('Doing normalize (nearest flats)')
tomo = tomopy.normalize_nf(projs, flat, dark, floc)
else:
logging.info('Doing normalize')
tomo = tomopy.normalize(projs, flat, dark)
#sinofilenametowrite = odirectory+'/rec'+iname[x]+'/'+iname[x]+'sino'
#dxchange.write_tiff_stack(tomo, fname=sinofilenametowrite, start=sinorange[0]+y*num_sino_per_chunk,axis=1)
projs = None
flat = None
logger.info("flat field correction..")
flat = np.median(flats, axis=0)
dark = np.median(darks, axis=0)
del flats, darks
# Only use middle of image
if crop_center is not None:
logger.info("Cropping to center of image")
slc = center_slices(projs[0], [crop_center, crop_center])
projs = projs[:, slc[0], slc[1]]
dark = dark[slc]
flat = flat[slc]
projs = tomopy.normalize(projs, flat[np.newaxis,:,:], dark[np.newaxis,:,:])
logger.info("Remove outliers")
tomopy.remove_outlier(projs, 0.1, out=projs)
# logger.info("shift correction using projections")
# shifts = proj_shift_correction(projs)
#print(shifts)
#np.savetxt("shifts_proj.txt", shifts)
#apply_shifts(projs, shifts, out=projs)
#shifted_projs = projs #change name after shifts
#del projs
#write_stack("shifted_projs_using_elastix", shifted_projs)
# initial shifts
#shifts = np.zeros((projs.shape[0], 2), np.float32)
#shifts = np.loadtxt("shifts_ltt.txt")
# # align using image correlation
def recon(
filename,
inputPath = './',
outputPath = None,
outputFilename = None,
doOutliers1D = False, # outlier removal in 1d (along sinogram columns)
outlier_diff1D = 750, # difference between good data and outlier data (outlier removal)
outlier_size1D = 3, # radius around each pixel to look for outliers (outlier removal)
doOutliers2D = False, # outlier removal, standard 2d on each projection
outlier_diff2D = 750, # difference between good data and outlier data (outlier removal)
outlier_size2D = 3, # radius around each pixel to look for outliers (outlier removal)
doFWringremoval = True, # Fourier-wavelet ring removal
doTIringremoval = False, # Titarenko ring removal
doSFringremoval = False, # Smoothing filter ring removal
ringSigma = 3, # damping parameter in Fourier space (Fourier-wavelet ring removal)
ringLevel = 8, # number of wavelet transform levels (Fourier-wavelet ring removal)
ringWavelet = 'db5', # type of wavelet filter (Fourier-wavelet ring removal)
ringNBlock = 0, # used in Titarenko ring removal (doTIringremoval)
ringAlpha = 1.5, # used in Titarenko ring removal (doTIringremoval)
ringSize = 5, # used in smoothing filter ring removal (doSFringremoval)
doPhaseRetrieval = False, # phase retrieval
alphaReg = 0.0002, # smaller = smoother (used for phase retrieval)
propagation_dist = 75, # sample-to-scintillator distance (phase retrieval)
kev = 24, # energy level (phase retrieval)
butterworth_cutoff = 0.25, #0.1 would be very smooth, 0.4 would be very grainy (reconstruction)
butterworth_order = 2, # for reconstruction
doTranslationCorrection = False, # correct for linear drift during scan
xshift = 0, # undesired dx transation correction (from 0 degree to 180 degree proj)
yshift = 0, # undesired dy transation correction (from 0 degree to 180 degree proj)
doPolarRing = False, # ring removal
Rarc=30, # min angle needed to be considered ring artifact (ring removal)
Rmaxwidth=100, # max width of rings to be filtered (ring removal)
Rtmax=3000.0, # max portion of image to filter (ring removal)
Rthr=3000.0, # max value of offset due to ring artifact (ring removal)
Rtmin=-3000.0, # min value of image to filter (ring removal)
cor=None, # center of rotation (float). If not used then cor will be detected automatically
corFunction = 'pc', # center of rotation function to use - can be 'pc', 'vo', or 'nm'
voInd = None, # index of slice to use for cor search (vo)
voSMin = -40, # min radius for searching in sinogram (vo)
voSMax = 40, # max radius for searching in sinogram (vo)
voSRad = 10, # search radius (vo)
voStep = 0.5, # search step (vo)
voRatio = 2.0, # ratio of field-of-view and object size (vo)
voDrop = 20, # drop lines around vertical center of mask (vo)
nmInd = None, # index of slice to use for cor search (nm)
nmInit = None, # initial guess for center (nm)
nmTol = 0.5, # desired sub-pixel accuracy (nm)
nmMask = True, # if True, limits analysis to circular region (nm)
nmRatio = 1.0, # ratio of radius of circular mask to edge of reconstructed image (nm)
nmSinoOrder = False, # if True, analyzes in sinogram space. If False, analyzes in radiograph space
use360to180 = False, # use 360 to 180 conversion
doBilateralFilter = False, # if True, uses bilateral filter on image just before write step # NOTE: image will be converted to 8bit if it is not already
bilateral_srad = 3, # spatial radius for bilateral filter (image will be converted to 8bit if not already)
bilateral_rrad = 30, # range radius for bilateral filter (image will be converted to 8bit if not already)
castTo8bit = False, # convert data to 8bit before writing
cast8bit_min=-10, # min value if converting to 8bit
cast8bit_max=30, # max value if converting to 8bit
useNormalize_nf = False, # normalize based on background intensity (nf)
chunk_proj = 100, # chunk size in projection direction
chunk_sino = 100, # chunk size in sinogram direction
npad = None, # amount to pad data before reconstruction
projused = None, #should be slicing in projection dimension (start,end,step)
sinoused = None, #should be sliceing in sinogram dimension (start,end,step). If first value is negative, it takes the number of slices from the second value in the middle of the stack.
correcttilt = 0, #tilt dataset
tiltcenter_slice = None, # tilt center (x direction)
tiltcenter_det = None, # tilt center (y direction)
angle_offset = 0, #this is the angle offset from our default (270) so that tomopy yields output in the same orientation as previous software (Octopus)
anglelist = None, #if not set, will assume evenly spaced angles which will be calculated by the angular range and number of angles found in the file. if set to -1, will read individual angles from each image. alternatively, a list of angles can be passed.
doBeamHardening = False, #turn on beam hardening correction, based on "Correction for beam hardening in computed tomography", Gabor Herman, 1979 Phys. Med. Biol. 24 81
BeamHardeningCoefficients = None, #6 values, tomo = a0 + a1*tomo + a2*tomo^2 + a3*tomo^3 + a4*tomo^4 + a5*tomo^5
projIgnoreList = None, #projections to be ignored in the reconstruction (for simplicity in the code, they will not be removed and will be processed as all other projections but will be set to zero absorption right before reconstruction.
*args, **kwargs):
start_time = time.time()
print("Start {} at:".format(filename)+time.strftime("%a, %d %b %Y %H:%M:%S +0000", time.localtime()))
outputPath = inputPath if outputPath is None else outputPath
outputFilename = filename if outputFilename is None else outputFilename
outputFilename = outputFilename.replace('.h5','')
tempfilenames = [outputPath+'tmp0.h5',outputPath+'tmp1.h5']
filenametowrite = outputPath+'/rec'+filename.strip(".h5")+'/'+outputFilename
#filenametowrite = outputPath+'/rec'+filename+'/'+outputFilename
print("cleaning up previous temp files", end="")
for tmpfile in tempfilenames:
try:
os.remove(tmpfile)
except OSError:
pass
print(", reading metadata")
datafile = h5py.File(inputPath+filename, 'r')
gdata = dict(dxchange.reader._find_dataset_group(datafile).attrs)
pxsize = float(gdata['pxsize'])/10 # /10 to convert units from mm to cm
numslices = int(gdata['nslices'])
numangles = int(gdata['nangles'])
angularrange = float(gdata['arange'])
numrays = int(gdata['nrays'])
npad = int(np.ceil(numrays * np.sqrt(2)) - numrays)//2 if npad is None else npad
projused = (0,numangles-1,1) if projused is None else projused
# ndark = int(gdata['num_dark_fields'])
# ind_dark = list(range(0, ndark))
# group_dark = [numangles - 1]
inter_bright = int(gdata['i0cycle'])
nflat = int(gdata['num_bright_field'])
ind_flat = list(range(0, nflat))
if inter_bright > 0:
group_flat = list(range(0, numangles, inter_bright))
if group_flat[-1] != numangles - 1:
group_flat.append(numangles - 1)
elif inter_bright == 0:
group_flat = [0, numangles - 1]
else:
group_flat = None
ind_tomo = list(range(0, numangles))
floc_independent = dxchange.reader._map_loc(ind_tomo, group_flat)
#figure out the angle list (a list of angles, one per projection image)
dtemp = datafile[list(datafile.keys())[0]]
fltemp = list(dtemp.keys())
firstangle = float(dtemp[fltemp[0]].attrs.get('rot_angle',0))
if anglelist is None:
#the offset angle should offset from the angle of the first image, which is usually 0, but in the case of timbir data may not be.
#we add the 270 to be inte same orientation as previous software used at bl832
angle_offset = 270 + angle_offset - firstangle
anglelist = tomopy.angles(numangles, angle_offset, angle_offset-angularrange)
elif anglelist==-1:
anglelist = np.zeros(shape=numangles)
for icount in range(0,numangles):
anglelist[icount] = np.pi/180*(270 + angle_offset - float(dtemp[fltemp[icount]].attrs['rot_angle']))
#if projused is different than default, need to chnage numangles and angularrange
#can't do useNormalize_nf and doOutliers2D at the same time, or doOutliers2D and doOutliers1D at the same time, b/c of the way we chunk, for now just disable that
if useNormalize_nf==True and doOutliers2D==True:
useNormalize_nf = False
print("we cannot currently do useNormalize_nf and doOutliers2D at the same time, turning off useNormalize_nf")
if doOutliers2D==True and doOutliers1D==True:
doOutliers1D = False
print("we cannot currently do doOutliers1D and doOutliers2D at the same time, turning off doOutliers1D")
#figure out how user can pass to do central x number of slices, or set of slices dispersed throughout (without knowing a priori the value of numslices)
if sinoused is None:
sinoused = (0,numslices,1)
elif sinoused[0]<0:
sinoused=(int(np.floor(numslices/2.0)-np.ceil(sinoused[1]/2.0)),int(np.floor(numslices/2.0)+np.floor(sinoused[1]/2.0)),1)
num_proj_per_chunk = np.minimum(chunk_proj,projused[1]-projused[0])
numprojchunks = (projused[1]-projused[0]-1)//num_proj_per_chunk+1
num_sino_per_chunk = np.minimum(chunk_sino,sinoused[1]-sinoused[0])
numsinochunks = (sinoused[1]-sinoused[0]-1)//num_sino_per_chunk+1
numprojused = (projused[1]-projused[0])//projused[2]
numsinoused = (sinoused[1]-sinoused[0])//sinoused[2]
BeamHardeningCoefficients = (0, 1, 0, 0, 0, .1) if BeamHardeningCoefficients is None else BeamHardeningCoefficients
if cor is None:
print("Detecting center of rotation", end="")
if angularrange>300:
lastcor = int(np.floor(numangles/2)-1)
else:
lastcor = numangles-1
#I don't want to see the warnings about the reader using a deprecated variable in dxchange
with warnings.catch_warnings():
warnings.simplefilter("ignore")
tomo, flat, dark, floc = dxchange.read_als_832h5(inputPath+filename,ind_tomo=(0,lastcor))
tomo = tomo.astype(np.float32)
if useNormalize_nf:
tomopy.normalize_nf(tomo, flat, dark, floc, out=tomo)
else:
tomopy.normalize(tomo, flat, dark, out=tomo)
if corFunction == 'vo':
# same reason for catching warnings as above
with warnings.catch_warnings():
warnings.simplefilter("ignore")
cor = tomopy.find_center_vo(tomo, ind=voInd, smin=voSMin, smax=voSMax, srad=voSRad, step=voStep,
ratio=voRatio, drop=voDrop)
elif corFunction == 'nm':
cor = tomopy.find_center(tomo, tomopy.angles(numangles, angle_offset, angle_offset-angularrange),
ind=nmInd, init=nmInit, tol=nmTol, mask=nmMask, ratio=nmRatio,
sinogram_order=nmSinoOrder)
elif corFunction == 'pc':
cor = tomopy.find_center_pc(tomo[0], tomo[1], tol=0.25)
else:
raise ValueError("\'corFunction\' must be one of: [ pc, vo, nm ].")
print(", {}".format(cor))
else:
print("using user input center of {}".format(cor))
function_list = []
if doOutliers1D:
function_list.append('remove_outlier1d')
if doOutliers2D:
function_list.append('remove_outlier2d')
if useNormalize_nf:
function_list.append('normalize_nf')
else:
function_list.append('normalize')
function_list.append('minus_log')
if doBeamHardening:
function_list.append('beam_hardening')
if doFWringremoval:
function_list.append('remove_stripe_fw')
if doTIringremoval:
function_list.append('remove_stripe_ti')
if doSFringremoval:
function_list.append('remove_stripe_sf')
if correcttilt:
function_list.append('correcttilt')
if use360to180:
function_list.append('do_360_to_180')
if doPhaseRetrieval:
function_list.append('phase_retrieval')
function_list.append('recon_mask')
if doPolarRing:
function_list.append('polar_ring')
if castTo8bit:
function_list.append('castTo8bit')
if doBilateralFilter:
function_list.append('bilateral_filter')
function_list.append('write_output')
# Figure out first direction to slice
for func in function_list:
if slice_dir[func] != 'both':
axis = slice_dir[func]
break
done = False
curfunc = 0
curtemp = 0
while True: # Loop over reading data in certain chunking direction
if axis=='proj':
niter = numprojchunks
else:
niter = numsinochunks
for y in range(niter): # Loop over chunks
print("{} chunk {} of {}".format(axis, y+1, niter))
if curfunc==0:
with warnings.catch_warnings():
warnings.simplefilter("ignore")
if axis=='proj':
tomo, flat, dark, floc = dxchange.read_als_832h5(inputPath+filename,ind_tomo=range(y*num_proj_per_chunk+projused[0],np.minimum((y + 1)*num_proj_per_chunk+projused[0],numangles)),sino=(sinoused[0],sinoused[1], sinoused[2]) )
else:
tomo, flat, dark, floc = dxchange.read_als_832h5(inputPath+filename,ind_tomo=range(projused[0],projused[1],projused[2]),sino=(y*num_sino_per_chunk+sinoused[0],np.minimum((y + 1)*num_sino_per_chunk+sinoused[0],numslices),1) )
else:
if axis=='proj':
start, end = y * num_proj_per_chunk, np.minimum((y + 1) * num_proj_per_chunk,numprojused)
tomo = dxchange.reader.read_hdf5(tempfilenames[curtemp],'/tmp/tmp',slc=((start,end,1),(0,numslices,1),(0,numrays,1))) #read in intermediate file
else:
start, end = y * num_sino_per_chunk, np.minimum((y + 1) * num_sino_per_chunk,numsinoused)
tomo = dxchange.reader.read_hdf5(tempfilenames[curtemp],'/tmp/tmp',slc=((0,numangles,1),(start,end,1),(0,numrays,1)))
dofunc = curfunc
keepvalues = None
while True: # Loop over operations to do in current chunking direction
func_name = function_list[dofunc]
newaxis = slice_dir[func_name]
if newaxis != 'both' and newaxis != axis:
# We have to switch axis, so flush to disk
if y==0:
try:
os.remove(tempfilenames[1-curtemp])
except OSError:
pass
appendaxis = 1 if axis=='sino' else 0
dxchange.writer.write_hdf5(tomo,fname=tempfilenames[1-curtemp],gname='tmp',dname='tmp',overwrite=False,appendaxis=appendaxis) #writing intermediate file...
break
print(func_name, end=" ")
curtime = time.time()
if func_name == 'remove_outlier1d':
tomo = tomo.astype(np.float32,copy=False)
remove_outlier1d(tomo, outlier_diff1D, size=outlier_size1D, out=tomo)
if func_name == 'remove_outlier2d':
tomo = tomo.astype(np.float32,copy=False)
tomopy.remove_outlier(tomo, outlier_diff2D, size=outlier_size2D, axis=0, out=tomo)
elif func_name == 'normalize_nf':
tomo = tomo.astype(np.float32,copy=False)
tomopy.normalize_nf(tomo, flat, dark, floc_independent, out=tomo) #use floc_independent b/c when you read file in proj chunks, you don't get the correct floc returned right now to use here.
elif func_name == 'normalize':
tomo = tomo.astype(np.float32,copy=False)
tomopy.normalize(tomo, flat, dark, out=tomo)
elif func_name == 'minus_log':
mx = np.float32(0.00000000000000000001)
ne.evaluate('where(tomo>mx, tomo, mx)', out=tomo)
tomopy.minus_log(tomo, out=tomo)
elif func_name == 'beam_hardening':
loc_dict = {'a{}'.format(i):np.float32(val) for i,val in enumerate(BeamHardeningCoefficients)}
tomo = ne.evaluate('a0 + a1*tomo + a2*tomo**2 + a3*tomo**3 + a4*tomo**4 + a5*tomo**5', local_dict=loc_dict, out=tomo)
elif func_name == 'remove_stripe_fw':
tomo = tomopy.remove_stripe_fw(tomo, sigma=ringSigma, level=ringLevel, pad=True, wname=ringWavelet)
elif func_name == 'remove_stripe_ti':
tomo = tomopy.remove_stripe_ti(tomo, nblock=ringNBlock, alpha=ringAlpha)
elif func_name == 'remove_stripe_sf':
tomo = tomopy.remove_stripe_sf(tomo, size=ringSize)
elif func_name == 'correcttilt':
if tiltcenter_slice is None:
tiltcenter_slice = numslices/2.
if tiltcenter_det is None:
tiltcenter_det = tomo.shape[2]/2
new_center = tiltcenter_slice - 0.5 - sinoused[0]
center_det = tiltcenter_det - 0.5
#add padding of 10 pixels, to be unpadded right after tilt correction. This makes the tilted image not have zeros at certain edges, which matters in cases where sample is bigger than the field of view. For the small amounts we are generally tilting the images, 10 pixels is sufficient.
# tomo = tomopy.pad(tomo, 2, npad=10, mode='edge')
# center_det = center_det + 10
cntr = (center_det, new_center)
for b in range(tomo.shape[0]):
tomo[b] = st.rotate(tomo[b], correcttilt, center=cntr, preserve_range=True, order=1, mode='edge', clip=True) #center=None means image is rotated around its center; order=1 is default, order of spline interpolation
# tomo = tomo[:, :, 10:-10]
elif func_name == 'do_360_to_180':
# Keep values around for processing the next chunk in the list
keepvalues = [angularrange, numangles, projused, num_proj_per_chunk, numprojchunks, numprojused, numrays, anglelist]
#why -.5 on one and not on the other?
if tomo.shape[0]%2>0:
tomo = sino_360_to_180(tomo[0:-1,:,:], overlap=int(np.round((tomo.shape[2]-cor-.5))*2), rotation='right')
angularrange = angularrange/2 - angularrange/(tomo.shape[0]-1)
else:
tomo = sino_360_to_180(tomo[:,:,:], overlap=int(np.round((tomo.shape[2]-cor))*2), rotation='right')
angularrange = angularrange/2
numangles = int(numangles/2)
projused = (0,numangles-1,1)
num_proj_per_chunk = np.minimum(chunk_proj,projused[1]-projused[0])
numprojchunks = (projused[1]-projused[0]-1)//num_proj_per_chunk+1
numprojused = (projused[1]-projused[0])//projused[2]
numrays = tomo.shape[2]
anglelist = anglelist[:numangles]
elif func_name == 'phase_retrieval':
tomo = tomopy.retrieve_phase(tomo, pixel_size=pxsize, dist=propagation_dist, energy=kev, alpha=alphaReg, pad=True)
elif func_name == 'translation_correction':
tomo = linear_translation_correction(tomo,dx=xshift,dy=yshift,interpolation=False):
elif func_name == 'recon_mask':
tomo = tomopy.pad(tomo, 2, npad=npad, mode='edge')
if projIgnoreList is not None:
for badproj in projIgnoreList:
tomo[badproj] = 0
rec = tomopy.recon(tomo, anglelist, center=cor+npad, algorithm='gridrec', filter_name='butterworth', filter_par=[butterworth_cutoff, butterworth_order])
rec = rec[:, npad:-npad, npad:-npad]
rec /= pxsize # convert reconstructed voxel values from 1/pixel to 1/cm
rec = tomopy.circ_mask(rec, 0)
elif func_name == 'polar_ring':
rec = np.ascontiguousarray(rec, dtype=np.float32)
rec = tomopy.remove_ring(rec, theta_min=Rarc, rwidth=Rmaxwidth, thresh_max=Rtmax, thresh=Rthr, thresh_min=Rtmin,out=rec)
elif func_name == 'castTo8bit':
rec = convert8bit(rec, cast8bit_min, cast8bit_max)
elif func_name == 'bilateral_filter':
rec = pyF3D.run_BilateralFilter(rec, spatialRadius=bilateral_srad, rangeRadius=bilateral_rrad)
elif func_name == 'write_output':
dxchange.write_tiff_stack(rec, fname=filenametowrite, start=y*num_sino_per_chunk + sinoused[0])
print('(took {:.2f} seconds)'.format(time.time()-curtime))
dofunc+=1
if dofunc==len(function_list):
break
if y<niter-1 and keepvalues: # Reset original values for next chunk
angularrange, numangles, projused, num_proj_per_chunk, numprojchunks, numprojused, numrays, anglelist = keepvalues
curtemp = 1 - curtemp
curfunc = dofunc
if curfunc==len(function_list):
break
axis = slice_dir[function_list[curfunc]]
print("cleaning up temp files")
for tmpfile in tempfilenames:
try:
os.remove(tmpfile)
except OSError:
pass
print("End Time: "+time.strftime("%a, %d %b %Y %H:%M:%S +0000", time.localtime()))
print('It took {:.3f} s to process {}'.format(time.time()-start_time,inputPath+filename))
def recon_hdf5(src_fanme,
dest_folder,
sino_range,
sino_step,
shift_grid,
center_vec=None,
center_eq=None,
dtype='float32',
algorithm='gridrec',
tolerance=1,
chunk_size=20,
save_sino=False,
sino_blur=None,
flattened_radius=120,
mode='180',
test_mode=False,
phase_retrieval=None,
ring_removal=True,
crop=None,
num_iter=None,
pad_length=0,
read_theta=True,
**kwargs):
"""
center_eq: a and b parameters in fitted center position equation center = a*slice + b.
"""
if not os.path.exists(dest_folder):
try:
os.mkdir(dest_folder)
except:
pass
sino_ini = int(sino_range[0])
sino_end = int(sino_range[1])
sino_ls_all = np.arange(sino_ini, sino_end, sino_step, dtype='int')
alloc_set = allocate_mpi_subsets(sino_ls_all.size,
size,
task_list=sino_ls_all)
sino_ls = alloc_set[rank]
# prepare metadata
f = h5py.File(src_fanme)
dset = f['exchange/data']
full_shape = dset.shape
if read_theta:
_, _, _, theta = read_data_adaptive(src_fanme, proj=(0, 1))
else:
theta = tomopy.angles(full_shape[0])
if center_eq is not None:
a, b = center_eq
center_ls = sino_ls_all * a + b
center_ls = np.round(center_ls)
for iblock in range(int(sino_ls.size / chunk_size) + 1):
internal_print('Beginning block {:d}.'.format(iblock))
t0 = time.time()
istart = iblock * chunk_size
iend = np.min([(iblock + 1) * chunk_size, sino_ls.size])
sub_sino_ls = sino_ls[istart:iend - 1]
center = np.take(center_ls, sub_sino_ls)
data = np.zeros([dset.shape[0], len(sub_sino_ls), dset.shape[2]])
for ind, i in enumerate(sub_sino_ls):
data[:, ind, :] = dset[:, i, :]
data[np.isnan(data)] = 0
data = data.astype('float32')
data = tomopy.remove_stripe_ti(data, alpha=4)
if sino_blur is not None:
for i in range(data.shape[1]):
data[:, i, :] = gaussian_filter(data[:, i, :], sino_blur)
if phase_retrieval:
data = tomopy.retrieve_phase(data, kwargs['pixel_size'],
kwargs['dist'], kwargs['energy'],
kwargs['alpha'])
if pad_length != 0:
data = pad_sinogram(data, pad_length)
data = tomopy.remove_stripe_ti(data, alpha=4)
if ring_removal:
rec0 = tomopy.recon(data,
theta,
center=center + pad_length,
algorithm=algorithm,
**kwargs)
rec = tomopy.remove_ring(np.copy(rec0))
cent = int((rec.shape[1] - 1) / 2)
xx, yy = np.meshgrid(np.arange(rec.shape[2]),
np.arange(rec.shape[1]))
mask0 = ((xx - cent)**2 +
(yy - cent)**2 <= flattened_radius**2)
mask = np.zeros(rec.shape, dtype='bool')
for i in range(mask.shape[0]):
mask[i, :, :] = mask0
rec[mask] = (rec[mask] + rec0[mask]) / 2
else:
rec = tomopy.recon(data,
theta,
center=center + pad_length,
algorithm=algorithm,
**kwargs)
if pad_length != 0:
rec = rec[:, pad_length:pad_length + full_shape[2],
pad_length:pad_length + full_shape[2]]
rec = tomopy.remove_outlier(rec, tolerance)
rec = tomopy.circ_mask(rec, axis=0, ratio=0.95)
if crop is not None:
crop = np.asarray(crop)
rec = rec[:, crop[0, 0]:crop[1, 0], crop[0, 1]:crop[1, 1]]
for i in range(rec.shape[0]):
slice = sub_sino_ls[i]
dxchange.write_tiff(
rec[i, :, :],
fname=os.path.join(
dest_folder, 'recon/recon_{:05d}.tiff').format(slice))
if save_sino:
dxchange.write_tiff(
data[:, i, :],
fname=os.path.join(
dest_folder, 'sino/recon_{:05d}_{:d}.tiff').format(
slice, int(center[i])))
iblock += 1
internal_print('Block {:d} finished in {:.2f} s.'.format(
iblock,
time.time() - t0))
else:
# divide chunks
grid_bins = np.append(np.ceil(shift_grid[:, 0, 0]), full_shape[1])
chunks = []
center_ls = []
istart = 0
counter = 0
# irow should be 0 for slice 0
irow = np.searchsorted(grid_bins, sino_ls[0], side='right') - 1
for i in range(sino_ls.size):
counter += 1
sino_next = i + 1 if i != sino_ls.size - 1 else i
if counter >= chunk_size or sino_ls[sino_next] >= grid_bins[
irow + 1] or sino_next == i:
iend = i + 1
chunks.append((istart, iend))
istart = iend
center_ls.append(center_vec[irow])
if sino_ls[sino_next] >= grid_bins[irow + 1]:
irow += 1
counter = 0
# reconstruct chunks
iblock = 1
for (istart, iend), center in zip(chunks, center_ls):
internal_print('Beginning block {:d}.'.format(iblock))
t0 = time.time()
internal_print('Reading data...')
sub_sino_ls = sino_ls[istart:iend]
data = np.zeros([dset.shape[0], len(sub_sino_ls), dset.shape[2]])
for ind, i in enumerate(sub_sino_ls):
data[:, ind, :] = dset[:, i, :]
if mode == '360':
overlap = 2 * (dset.shape[2] - center)
data = tomosaic.sino_360_to_180(data,
overlap=overlap,
rotation='right')
theta = tomopy.angles(data.shape[0])
data[np.isnan(data)] = 0
data = data.astype('float32')
if sino_blur is not None:
for i in range(data.shape[1]):
data[:, i, :] = gaussian_filter(data[:, i, :], sino_blur)
if phase_retrieval:
data = tomopy.retrieve_phase(data, kwargs['pixel_size'],
kwargs['dist'], kwargs['energy'],
kwargs['alpha'])
if pad_length != 0:
data = pad_sinogram(data, pad_length)
data = tomopy.remove_stripe_ti(data, alpha=4)
if ring_removal:
rec0 = tomopy.recon(data,
theta,
center=center + pad_length,
algorithm=algorithm,
**kwargs)
rec = tomopy.remove_ring(np.copy(rec0))
cent = int((rec.shape[1] - 1) / 2)
xx, yy = np.meshgrid(np.arange(rec.shape[2]),
np.arange(rec.shape[1]))
mask0 = ((xx - cent)**2 +
(yy - cent)**2 <= flattened_radius**2)
mask = np.zeros(rec.shape, dtype='bool')
for i in range(mask.shape[0]):
mask[i, :, :] = mask0
rec[mask] = (rec[mask] + rec0[mask]) / 2
else:
rec = tomopy.recon(data,
theta,
center=center + pad_length,
algorithm=algorithm,
**kwargs)
if pad_length != 0:
rec = rec[:, pad_length:pad_length + full_shape[2],
pad_length:pad_length + full_shape[2]]
rec = tomopy.remove_outlier(rec, tolerance)
rec = tomopy.circ_mask(rec, axis=0, ratio=0.95)
if crop is not None:
crop = np.asarray(crop)
rec = rec[:, crop[0, 0]:crop[1, 0], crop[0, 1]:crop[1, 1]]
for i in range(rec.shape[0]):
slice = sub_sino_ls[i]
if test_mode:
dxchange.write_tiff(
rec[i, :, :],
fname=os.path.join(
dest_folder,
'recon/recon_{:05d}_{:d}.tiff').format(
slice, center),
dtype=dtype)
else:
dxchange.write_tiff(
rec[i, :, :],
fname=os.path.join(
dest_folder,
'recon/recon_{:05d}.tiff').format(slice),
dtype=dtype)
if save_sino:
dxchange.write_tiff(
data[:, i, :],
fname=os.path.join(
dest_folder, 'sino/recon_{:05d}_{:d}.tiff').format(
slice, center),
dtype=dtype)
internal_print('Block {:d} finished in {:.2f} s.'.format(
iblock,
time.time() - t0))
iblock += 1
return
def recon_hdf5_mpi(src_fanme,
dest_folder,
sino_range,
sino_step,
center_vec,
shift_grid,
dtype='float32',
algorithm='gridrec',
tolerance=1,
save_sino=False,
sino_blur=None,
**kwargs):
"""
Reconstruct a single tile, or fused HDF5 created using util/total_fusion. MPI supported.
"""
raise DeprecationWarning
if rank == 0:
if not os.path.exists(dest_folder):
os.mkdir(dest_folder)
sino_ini = int(sino_range[0])
sino_end = int(sino_range[1])
f = h5py.File(src_fanme)
dset = f['exchange/data']
full_shape = dset.shape
theta = tomopy.angles(full_shape[0])
center_vec = np.asarray(center_vec)
sino_ls = np.arange(sino_ini, sino_end, sino_step, dtype='int')
grid_bins = np.ceil(shift_grid[:, 0, 0])
t0 = time.time()
alloc_set = allocate_mpi_subsets(sino_ls.size, size, task_list=sino_ls)
for slice in alloc_set[rank]:
print(' Rank {:d}: reconstructing {:d}'.format(rank, slice))
grid_line = np.digitize(slice, grid_bins)
grid_line = grid_line - 1
center = center_vec[grid_line]
data = dset[:, slice, :]
if sino_blur is not None:
data = gaussian_filter(data, sino_blur)
data = data.reshape([full_shape[0], 1, full_shape[2]])
data[np.isnan(data)] = 0
data = data.astype('float32')
if save_sino:
dxchange.write_tiff(data[:, slice, :],
fname=os.path.join(
dest_folder,
'sino/recon_{:05d}_{:d}.tiff').format(
slice, center))
# data = tomopy.remove_stripe_ti(data)
rec = tomopy.recon(data,
theta,
center=center,
algorithm=algorithm,
**kwargs)
# rec = tomopy.remove_ring(rec)
rec = tomopy.remove_outlier(rec, tolerance)
rec = tomopy.circ_mask(rec, axis=0, ratio=0.95)
dxchange.write_tiff(rec,
fname='{:s}/recon/recon_{:05d}_{:d}'.format(
dest_folder, slice, center),
dtype=dtype)
print('Rank {:d} finished in {:.2f} s.'.format(rank, time.time() - t0))
return
