def create_preprocessing_pipeline(args,
graph,
source=None,
processing_node=None):
pm = Ufo.PluginManager()
if not (args.width and args.height):
width, height = determine_shape(args, args.projections)
if not width:
raise RuntimeError("Could not determine width from the input")
if not args.width:
args.width = width
if not args.height:
args.height = height - args.y
LOG.debug('Image width x height: %d x %d', args.width, args.height)
if source:
current = source
elif args.darks and args.flats:
current = create_flat_correct_pipeline(args,
graph,
processing_node=processing_node)
else:
current = get_task('read')
set_node_props(current, args)
if not args.projections:
raise RuntimeError('--projections not set')
setup_read_task(current, args.projections, args)
if args.absorptivity:
absorptivity = get_task('calculate',
processing_node=processing_node)
absorptivity.props.expression = '-log(v)'
graph.connect_nodes(current, absorptivity)
current = absorptivity
if args.transpose_input:
transpose = get_task('transpose')
graph.connect_nodes(current, transpose)
current = transpose
tmp = args.width
args.width = args.height
args.height = tmp
if args.projection_filter != 'none':
pf_first, pf_last = create_projection_filtering_pipeline(
args, graph, processing_node=processing_node)
graph.connect_nodes(current, pf_first)
current = pf_last
if args.energy is not None and args.propagation_distance is not None:
pr_first, pr_last = create_phase_retrieval_pipeline(
args, graph, processing_node=processing_node)
graph.connect_nodes(current, pr_first)
current = pr_last
return current
def get_sinogram_reader(params):
reader = get_file_reader(params)
setup_read_task(reader, params.sinograms, params)
image = read_image(get_first_filename(params.sinograms))
if len(image.shape) > 2:
# this is a probably a multi TIFF/raw
width, height = image.shape[2], image.shape[1]
else:
# this is a directory of sinograms
width, height = image.shape[1], image.shape[0]
return reader, width, height
def get_sinogram_reader(params):
reader = get_file_reader(params)
setup_read_task(reader, params.sinograms, params)
image = read_image(get_first_filename(params.sinograms))
if len(image.shape) > 2:
# this is a probably a multi TIFF/raw
width, height = image.shape[2], image.shape[1]
else:
# this is a directory of sinograms
width, height = image.shape[1], image.shape[0]
return reader, width, height
def _setup_source(params, pm, graph):
from tofu.preprocess import create_flat_correct_pipeline
from tofu.util import set_node_props, setup_read_task
if params.dry_run:
source = pm.get_task('dummy-data')
source.props.number = params.number
source.props.width = params.width
source.props.height = params.height
elif params.darks and params.flats:
source = create_flat_correct_pipeline(params, graph)
else:
source = pm.get_task('read')
set_node_props(source, params)
setup_read_task(source, params.projections, params)
return source
def get_projection_reader(params):
reader = get_file_reader(params)
setup_read_task(reader, params.projections, params)
width, height = determine_shape(params, params.projections)
return reader, width, height
def get_projection_reader(params):
reader = get_file_reader(params)
setup_read_task(reader, params.projections, params)
width, height = determine_shape(params, params.projections)
return reader, width, height
def create_flat_correct_pipeline(args, graph):
"""
Create flat field correction pipeline. All the settings are provided in
*args*. *graph* is used for making the connections. Returns the flat field
correction task which can be used for further pipelining.
"""
pm = Ufo.PluginManager()
if args.projections is None or args.flats is None or args.darks is None:
raise RuntimeError(
"You must specify --projections, --flats and --darks.")
def get_task(name, **kwargs):
"""Get task *name* with properties *kwargs*."""
task = pm.get_task(name)
task.set_properties(**kwargs)
return task
reader = get_task('read')
dark_reader = get_task('read')
flat_before_reader = get_task('read')
ffc = get_task('flat-field-correct',
dark_scale=args.dark_scale,
absorption_correct=args.absorptivity,
fix_nan_and_inf=args.fix_nan_and_inf)
mode = args.reduction_mode.lower()
roi_args = make_subargs(args, ['y', 'height', 'y_step'])
set_node_props(reader, args)
set_node_props(dark_reader, roi_args)
set_node_props(flat_before_reader, roi_args)
for r, path in ((reader, args.projections), (dark_reader, args.darks),
(flat_before_reader, args.flats)):
setup_read_task(r, path, args)
LOG.debug(
"Doing flat field correction using reduction mode `{}'".format(mode))
if args.flats2:
flat_after_reader = get_task('read')
setup_read_task(flat_after_reader, args.flats2, args)
set_node_props(flat_after_reader, roi_args)
num_files = len(get_filenames(args.projections))
can_read = len(range(args.start, num_files, args.step))
number = args.number if args.number else num_files
num_read = min(can_read, number)
flat_interpolate = get_task('interpolate', number=num_read)
if args.resize:
LOG.debug("Resize input data by factor of {}".format(args.resize))
proj_bin = get_task('bin', size=args.resize)
dark_bin = get_task('bin', size=args.resize)
flat_bin = get_task('bin', size=args.resize)
graph.connect_nodes(reader, proj_bin)
graph.connect_nodes(dark_reader, dark_bin)
graph.connect_nodes(flat_before_reader, flat_bin)
reader, dark_reader, flat_before_reader = proj_bin, dark_bin, flat_bin
if args.flats2:
flat_bin = get_task('bin', size=args.resize)
graph.connect_nodes(flat_after_reader, flat_bin)
flat_after_reader = flat_bin
if mode == 'median':
dark_stack = get_task('stack', number=len(get_filenames(args.darks)))
dark_reduced = get_task('flatten', mode='median')
flat_before_stack = get_task('stack',
number=len(get_filenames(args.flats)))
flat_before_reduced = get_task('flatten', mode='median')
graph.connect_nodes(dark_reader, dark_stack)
graph.connect_nodes(dark_stack, dark_reduced)
graph.connect_nodes(flat_before_reader, flat_before_stack)
graph.connect_nodes(flat_before_stack, flat_before_reduced)
if args.flats2:
flat_after_stack = get_task('stack',
number=len(get_filenames(args.flats2)))
flat_after_reduced = get_task('flatten', mode='median')
graph.connect_nodes(flat_after_reader, flat_after_stack)
graph.connect_nodes(flat_after_stack, flat_after_reduced)
elif mode == 'average':
dark_reduced = get_task('average')
flat_before_reduced = get_task('average')
graph.connect_nodes(dark_reader, dark_reduced)
graph.connect_nodes(flat_before_reader, flat_before_reduced)
if args.flats2:
flat_after_reduced = get_task('average')
graph.connect_nodes(flat_after_reader, flat_after_reduced)
else:
raise ValueError('Invalid reduction mode')
graph.connect_nodes_full(reader, ffc, 0)
graph.connect_nodes_full(dark_reduced, ffc, 1)
if args.flats2:
graph.connect_nodes_full(flat_before_reduced, flat_interpolate, 0)
graph.connect_nodes_full(flat_after_reduced, flat_interpolate, 1)
graph.connect_nodes_full(flat_interpolate, ffc, 2)
else:
graph.connect_nodes_full(flat_before_reduced, ffc, 2)
return ffc
def create_flat_correct_pipeline(args, graph):
"""
Create flat field correction pipeline. All the settings are provided in
*args*. *graph* is used for making the connections. Returns the flat field
correction task which can be used for further pipelining.
"""
pm = Ufo.PluginManager()
if args.projections is None or args.flats is None or args.darks is None:
raise RuntimeError("You must specify --projections, --flats and --darks.")
def get_task(name, **kwargs):
"""Get task *name* with properties *kwargs*."""
task = pm.get_task(name)
task.set_properties(**kwargs)
return task
reader = get_task('read')
dark_reader = get_task('read')
flat_before_reader = get_task('read')
ffc = get_task('flat-field-correct', dark_scale=args.dark_scale,
absorption_correct=args.absorptivity,
fix_nan_and_inf=args.fix_nan_and_inf)
mode = args.reduction_mode.lower()
roi_args = make_subargs(args, ['y', 'height', 'y_step'])
set_node_props(reader, args)
set_node_props(dark_reader, roi_args)
set_node_props(flat_before_reader, roi_args)
for r, path in ((reader, args.projections), (dark_reader, args.darks), (flat_before_reader, args.flats)):
setup_read_task(r, path, args)
LOG.debug("Doing flat field correction using reduction mode `{}'".format(mode))
if args.flats2:
flat_after_reader = get_task('read')
setup_read_task(flat_after_reader, args.flats2, args)
set_node_props(flat_after_reader, roi_args)
num_files = len(get_filenames(args.projections))
can_read = len(range(args.start, num_files, args.step))
number = args.number if args.number else num_files
num_read = min(can_read, number)
flat_interpolate = get_task('interpolate', number=num_read)
if args.resize:
LOG.debug("Resize input data by factor of {}".format(args.resize))
proj_bin = get_task('bin', size=args.resize)
dark_bin = get_task('bin', size=args.resize)
flat_bin = get_task('bin', size=args.resize)
graph.connect_nodes(reader, proj_bin)
graph.connect_nodes(dark_reader, dark_bin)
graph.connect_nodes(flat_before_reader, flat_bin)
reader, dark_reader, flat_before_reader = proj_bin, dark_bin, flat_bin
if args.flats2:
flat_bin = get_task('bin', size=args.resize)
graph.connect_nodes(flat_after_reader, flat_bin)
flat_after_reader = flat_bin
if mode == 'median':
dark_stack = get_task('stack', number=len(get_filenames(args.darks)))
dark_reduced = get_task('flatten', mode='median')
flat_before_stack = get_task('stack', number=len(get_filenames(args.flats)))
flat_before_reduced = get_task('flatten', mode='median')
graph.connect_nodes(dark_reader, dark_stack)
graph.connect_nodes(dark_stack, dark_reduced)
graph.connect_nodes(flat_before_reader, flat_before_stack)
graph.connect_nodes(flat_before_stack, flat_before_reduced)
if args.flats2:
flat_after_stack = get_task('stack', number=len(get_filenames(args.flats2)))
flat_after_reduced = get_task('flatten', mode='median')
graph.connect_nodes(flat_after_reader, flat_after_stack)
graph.connect_nodes(flat_after_stack, flat_after_reduced)
elif mode == 'average':
dark_reduced = get_task('average')
flat_before_reduced = get_task('average')
graph.connect_nodes(dark_reader, dark_reduced)
graph.connect_nodes(flat_before_reader, flat_before_reduced)
if args.flats2:
flat_after_reduced = get_task('average')
graph.connect_nodes(flat_after_reader, flat_after_reduced)
else:
raise ValueError('Invalid reduction mode')
graph.connect_nodes_full(reader, ffc, 0)
graph.connect_nodes_full(dark_reduced, ffc, 1)
if args.flats2:
graph.connect_nodes_full(flat_before_reduced, flat_interpolate, 0)
graph.connect_nodes_full(flat_after_reduced, flat_interpolate, 1)
graph.connect_nodes_full(flat_interpolate, ffc, 2)
else:
graph.connect_nodes_full(flat_before_reduced, ffc, 2)
return ffc
def create_preprocessing_pipeline(args, graph, source=None, processing_node=None,
cone_beam_weight=True, make_reader=True):
"""If *make_reader* is True, create a read task if *source* is None and no dark and flat fields
are given.
"""
import numpy as np
if not (args.width and args.height):
width, height = determine_shape(args, args.projections)
if not width:
raise RuntimeError("Could not determine width from the input")
if not args.width:
args.width = width
if not args.height:
args.height = height - args.y
LOG.debug('Image width x height: %d x %d', args.width, args.height)
current = None
if source:
current = source
elif args.darks and args.flats:
current = create_flat_correct_pipeline(args, graph, processing_node=processing_node)
else:
if make_reader:
current = get_task('read')
set_node_props(current, args)
if not args.projections:
raise RuntimeError('--projections not set')
setup_read_task(current, args.projections, args)
if args.absorptivity:
absorptivity = get_task('calculate', processing_node=processing_node)
absorptivity.props.expression = 'v <= 0 ? 0.0f : -log(v)'
if current:
graph.connect_nodes(current, absorptivity)
current = absorptivity
if args.transpose_input:
transpose = get_task('transpose')
if current:
graph.connect_nodes(current, transpose)
current = transpose
tmp = args.width
args.width = args.height
args.height = tmp
if cone_beam_weight and not np.all(np.isinf(args.source_position_y)):
# Cone beam projection weight
LOG.debug('Enabling cone beam weighting')
weight = get_task('cone-beam-projection-weight', processing_node=processing_node)
weight.props.source_distance = (-np.array(args.source_position_y)).tolist()
weight.props.detector_distance = args.detector_position_y
weight.props.center_position_x = args.center_position_x or [args.width / 2. + (args.width % 2) * 0.5]
weight.props.center_position_z = args.center_position_z or [args.height / 2. + (args.height % 2) * 0.5]
weight.props.axis_angle_x = args.axis_angle_x
if current:
graph.connect_nodes(current, weight)
current = weight
if args.energy is not None and args.propagation_distance is not None:
pr_first, pr_last = create_phase_retrieval_pipeline(args, graph,
processing_node=processing_node)
if current:
graph.connect_nodes(current, pr_first)
current = pr_last
if args.projection_filter != 'none':
pf_first, pf_last = create_projection_filtering_pipeline(args, graph,
processing_node=processing_node)
if current:
graph.connect_nodes(current, pf_first)
current = pf_last
return current
def create_preprocessing_pipeline(args,
graph,
source=None,
processing_node=None,
cone_beam_weight=True,
make_reader=True):
"""If *make_reader* is True, create a read task if *source* is None and no dark and flat fields
are given.
"""
import numpy as np
if not (args.width and args.height):
width, height = determine_shape(args, args.projections)
if not width:
raise RuntimeError("Could not determine width from the input")
if not args.width:
args.width = width
if not args.height:
args.height = height - args.y
LOG.debug('Image width x height: %d x %d', args.width, args.height)
current = None
if source:
current = source
elif args.darks and args.flats:
current = create_flat_correct_pipeline(args,
graph,
processing_node=processing_node)
else:
if make_reader:
current = get_task('read')
set_node_props(current, args)
if not args.projections:
raise RuntimeError('--projections not set')
setup_read_task(current, args.projections, args)
if args.absorptivity:
absorptivity = get_task('calculate',
processing_node=processing_node)
absorptivity.props.expression = 'v <= 0 ? 0.0f : -log(v)'
if current:
graph.connect_nodes(current, absorptivity)
current = absorptivity
if args.transpose_input:
transpose = get_task('transpose')
if current:
graph.connect_nodes(current, transpose)
current = transpose
tmp = args.width
args.width = args.height
args.height = tmp
if cone_beam_weight and not np.all(np.isinf(args.source_position_y)):
# Cone beam projection weight
LOG.debug('Enabling cone beam weighting')
weight = get_task('cone-beam-projection-weight',
processing_node=processing_node)
weight.props.source_distance = (
-np.array(args.source_position_y)).tolist()
weight.props.detector_distance = args.detector_position_y
weight.props.center_position_x = args.center_position_x or [
args.width / 2. + (args.width % 2) * 0.5
]
weight.props.center_position_z = args.center_position_z or [
args.height / 2. + (args.height % 2) * 0.5
]
weight.props.axis_angle_x = args.axis_angle_x
if current:
graph.connect_nodes(current, weight)
current = weight
if args.energy is not None and args.propagation_distance is not None:
pr_first, pr_last = create_phase_retrieval_pipeline(
args, graph, processing_node=processing_node)
if current:
graph.connect_nodes(current, pr_first)
current = pr_last
if args.projection_filter != 'none':
pf_first, pf_last = create_projection_filtering_pipeline(
args, graph, processing_node=processing_node)
if current:
graph.connect_nodes(current, pf_first)
current = pf_last
return current