def load_projections(self):
'''
Load projections and return AcquisitionData container
'''
# get path to projections
path_projection = os.path.dirname(self.xtek_file)
# get number of projections
num_projections = numpy.shape(self._ag.angles)[0]
# allocate array to store projections
data = numpy.zeros(
(num_projections, self._ag.pixel_num_v, self._ag.pixel_num_h),
dtype=float)
for i in range(num_projections):
filename = (path_projection + '/' + self._experiment_name +
'_{:04d}.tif').format(i + 1)
try:
tmp = numpy.asarray(Image.open(filename), dtype=float)
except:
print('Error reading\n {}\n file.'.format(filename))
raise
if (self.binning == [1, 1]):
data[i, :, :] = tmp[self._roi_par[0][0]:self._roi_par[0][1],
self._roi_par[1][0]:self._roi_par[1][1]]
else:
shape = (self._ag.pixel_num_v, self.binning[0],
self._ag.pixel_num_h, self.binning[1])
data[i, :, :] = tmp[self._roi_par[0][0]:(self._roi_par[0][0] + (((self._roi_par[0][1] - self._roi_par[0][0]) // self.binning[0]) * self.binning[0])), \
self._roi_par[1][0]:(self._roi_par[1][0] + (((self._roi_par[1][1] - self._roi_par[1][0]) // self.binning[1]) * self.binning[1]))].reshape(shape).mean(-1).mean(1)
if (self.normalize):
data /= self._white_level
data[data > 1] = 1
if self.flip:
return AcquisitionData(array = data[:, :, ::-1],
deep_copy = False,
geometry = self._ag,
dimension_labels = ['angle', \
'vertical', \
'horizontal'])
else:
return AcquisitionData(array = data,
deep_copy = False,
geometry = self._ag,
dimension_labels = ['angle', \
'vertical', \
'horizontal'])
def process(self):
IM = self.get_input()
DATA = AcquisitionData(geometry=self.sinogram_geometry,
dimension_labels=self.output_axes_order)
sinogram_id, DATA.array = astra.create_sino3d_gpu(
IM.as_array(), self.proj_geom, self.vol_geom)
astra.data3d.delete(sinogram_id)
# 3D CUDA FP does not need scaling
return DATA
def allocate_direct(self):
if issubclass(type(self.geometry), ImageGeometry):
return ImageData(geometry=self.geometry)
elif issubclass(type(self.geometry), AcquisitionGeometry):
return AcquisitionData(geometry=self.geometry)
else:
raise ValueError(
"Wrong geometry type: expected ImageGeometry of AcquisitionGeometry, got ",
type(self.geometry))
def test_AcquisitionData(self):
sgeometry = AcquisitionGeometry(dimension=2,
angles=numpy.linspace(0, 180, num=10),
geom_type='parallel',
pixel_num_v=3,
pixel_num_h=5,
channels=2)
sino = AcquisitionData(geometry=sgeometry)
self.assertEqual(sino.shape, (2, 10, 3, 5))
def process(self):
IM = self.get_input()
#create the output AcquisitionData
DATA = AcquisitionData(geometry=self.sinogram_geometry)
for k in range(DATA.geometry.channels):
sinogram_id, DATA.as_array()[k] = astra.create_sino(
IM.as_array()[k], self.proj_id)
astra.data2d.delete(sinogram_id)
if self.device == 'cpu':
return DATA
else:
if self.sinogram_geometry.geom_type == 'cone':
return DATA
else:
scaling = (1.0 / self.volume_geometry.voxel_size_x)
return scaling * DATA
def setupCCPiGeometries(voxel_num_x, voxel_num_y, voxel_num_z, angles, counter):
vg = ImageGeometry(voxel_num_x=voxel_num_x,voxel_num_y=voxel_num_y, voxel_num_z=voxel_num_z)
Phantom_ccpi = ImageData(geometry=vg,
dimension_labels=['horizontal_x','horizontal_y','vertical'])
#.subset(['horizontal_x','horizontal_y','vertical'])
# ask the ccpi code what dimensions it would like
voxel_per_pixel = 1
geoms = pbalg.pb_setup_geometry_from_image(Phantom_ccpi.as_array(),
angles,
voxel_per_pixel )
pg = AcquisitionGeometry('parallel',
'3D',
angles,
geoms['n_h'], 1.0,
geoms['n_v'], 1.0 #2D in 3D is a slice 1 pixel thick
)
center_of_rotation = Phantom_ccpi.get_dimension_size('horizontal_x') / 2
ad = AcquisitionData(geometry=pg,dimension_labels=['angle','vertical','horizontal'])
geoms_i = pbalg.pb_setup_geometry_from_acquisition(ad.as_array(),
angles,
center_of_rotation,
voxel_per_pixel )
counter+=1
if counter < 4:
if (not ( geoms_i == geoms )):
print ("not equal and {0}".format(counter))
X = max(geoms['output_volume_x'], geoms_i['output_volume_x'])
Y = max(geoms['output_volume_y'], geoms_i['output_volume_y'])
Z = max(geoms['output_volume_z'], geoms_i['output_volume_z'])
return setupCCPiGeometries(X,Y,Z,angles, counter)
else:
return geoms
else:
return geoms_i
def process(self):
IM = self.get_input()
DATA = AcquisitionData(geometry=self.sinogram_geometry)
#sinogram_id, DATA = astra.create_sino( IM.as_array(),
# self.proj_id)
sinogram_id, DATA.array = astra.create_sino(IM.as_array(),
self.proj_id)
astra.data2d.delete(sinogram_id)
if self.device == 'cpu':
return DATA
else:
if self.sinogram_geometry.geom_type == 'cone':
return DATA
else:
scaling = 1.0 / self.volume_geometry.voxel_size_x
return scaling * DATA
def get_acquisition_data(self, dimensions=None):
'''
This method load the acquisition data and given dimension and returns an AcquisitionData Object
'''
data = self.load_projection(dimensions)
dims = self.get_projection_dimensions()
geometry = AcquisitionGeometry('parallel',
'3D',
self.get_projection_angles(),
pixel_num_h=dims[2],
pixel_size_h=1,
pixel_num_v=dims[1],
pixel_size_v=1,
dist_source_center=None,
dist_center_detector=None,
channels=1)
return AcquisitionData(
data,
geometry=geometry,
dimension_labels=['angle', 'vertical', 'horizontal'])
def test_AcquisitionGeometry_allocate(self):
ageometry = AcquisitionGeometry(dimension=2,
angles=numpy.linspace(0, 180, num=10),
geom_type='parallel', pixel_num_v=3,
pixel_num_h=5, channels=2)
sino = ageometry.allocate()
shape = sino.shape
print ("shape", shape)
self.assertEqual(0,sino.as_array()[0][0][0][0])
self.assertEqual(0,sino.as_array()[shape[0]-1][shape[1]-1][shape[2]-1][shape[3]-1])
sino = ageometry.allocate(1)
self.assertEqual(1,sino.as_array()[0][0][0][0])
self.assertEqual(1,sino.as_array()[shape[0]-1][shape[1]-1][shape[2]-1][shape[3]-1])
print (sino.dimension_labels, sino.shape, ageometry)
default_order = ['channel' , 'angle' ,
'vertical' , 'horizontal']
self.assertEqual(default_order[0], sino.dimension_labels[0])
self.assertEqual(default_order[1], sino.dimension_labels[1])
self.assertEqual(default_order[2], sino.dimension_labels[2])
self.assertEqual(default_order[3], sino.dimension_labels[3])
order = ['vertical' , 'horizontal', 'channel' , 'angle' ]
sino = ageometry.allocate(0,dimension_labels=order)
print (sino.dimension_labels, sino.shape, ageometry)
self.assertEqual(order[0], sino.dimension_labels[0])
self.assertEqual(order[1], sino.dimension_labels[1])
self.assertEqual(order[2], sino.dimension_labels[2])
self.assertEqual(order[2], sino.dimension_labels[2])
try:
z = AcquisitionData(numpy.random.randint(10, size=(2,3)), geometry=ageometry)
self.assertTrue(False)
except ValueError as ve:
print (ve)
self.assertTrue(True)
def process(self):
volume = self.get_input()
volume_axes = volume.get_data_axes_order(new_order=self.default_image_axes_order)
if not volume_axes == [0,1,2]:
volume.array = numpy.transpose(volume.array, volume_axes)
pixel_per_voxel = 1 # should be estimated from image_geometry and
# acquisition_geometry
if self.acquisition_geometry.geom_type == 'parallel':
pixels = pbalg.pb_forward_project(volume.as_array(),
self.acquisition_geometry.angles,
pixel_per_voxel)
out = AcquisitionData(geometry=self.acquisition_geometry,
label_dimensions=self.default_acquisition_axes_order)
out.fill(pixels)
out_axes = out.get_data_axes_order(new_order=self.output_axes_order)
if not out_axes == [0,1,2]:
out.array = numpy.transpose(out.array, out_axes)
return out
else:
raise ValueError('Cannot process cone beam')
def get_acquisition_data(self, dimensions=None):
'''
This method load the acquisition data and given dimension and returns an AcquisitionData Object
'''
data = self.load_projection(dimensions)
return AcquisitionData(data, geometry=self.geometry)
def get_acquisition_data_batch(self, bmin=None, bmax=None):
if not h5pyAvailable:
raise Exception("Error: h5py is not installed")
if self.filename is None:
return
try:
with NexusFile(self.filename, 'r') as file:
try:
dims = self.get_projection_dimensions()
except KeyError:
dims = file[self.data_path].shape
if bmin is None or bmax is None:
raise ValueError(
'get_acquisition_data_batch: please specify fastest index batch limits'
)
if bmin >= 0 and bmin < bmax and bmax <= dims[0]:
data = np.array(file[self.data_path][bmin:bmax])
else:
raise ValueError(
'get_acquisition_data_batch: bmin {0}>0 bmax {1}<{2}'.
format(bmin, bmax, dims[0]))
except:
print("Error reading nexus file")
raise
try:
angles = self.get_projection_angles()[bmin:bmax]
except KeyError as ke:
n = data.shape[0]
angles = np.linspace(0, n, n + 1, dtype=np.float32)[bmin:bmax]
if bmax - bmin > 1:
geometry = AcquisitionGeometry('parallel',
'3D',
angles,
pixel_num_h=dims[2],
pixel_size_h=1,
pixel_num_v=bmax - bmin,
pixel_size_v=1,
dist_source_center=None,
dist_center_detector=None,
channels=1)
return AcquisitionData(
data,
False,
geometry=geometry,
dimension_labels=['angle', 'vertical', 'horizontal'])
elif bmax - bmin == 1:
geometry = AcquisitionGeometry('parallel',
'2D',
angles,
pixel_num_h=dims[2],
pixel_size_h=1,
dist_source_center=None,
dist_center_detector=None,
channels=1)
return AcquisitionData(data.squeeze(),
False,
geometry=geometry,
dimension_labels=['angle', 'horizontal'])
def get_acquisition_data_subset(self, ymin=None, ymax=None):
'''
This method load the acquisition data and given dimension and returns an AcquisitionData Object
'''
if not h5pyAvailable:
raise Exception("Error: h5py is not installed")
if self.filename is None:
return
try:
with NexusFile(self.filename, 'r') as file:
try:
dims = self.get_projection_dimensions()
except KeyError:
pass
dims = file[self.data_path].shape
if ymin is None and ymax is None:
data = np.array(file[self.data_path])
else:
if ymin is None:
ymin = 0
if ymax > dims[1]:
raise ValueError('ymax out of range')
data = np.array(file[self.data_path][:, :ymax, :])
elif ymax is None:
ymax = dims[1]
if ymin < 0:
raise ValueError('ymin out of range')
data = np.array(file[self.data_path][:, ymin:, :])
else:
if ymax > dims[1]:
raise ValueError('ymax out of range')
if ymin < 0:
raise ValueError('ymin out of range')
data = np.array(file[self.data_path][:, ymin:ymax, :])
except:
print("Error reading nexus file")
raise
try:
angles = self.get_projection_angles()
except KeyError as ke:
n = data.shape[0]
angles = np.linspace(0, n, n + 1, dtype=np.float32)
if ymax - ymin > 1:
geometry = AcquisitionGeometry('parallel',
'3D',
angles,
pixel_num_h=dims[2],
pixel_size_h=1,
pixel_num_v=ymax - ymin,
pixel_size_v=1,
dist_source_center=None,
dist_center_detector=None,
channels=1)
return AcquisitionData(
data,
False,
geometry=geometry,
dimension_labels=['angle', 'vertical', 'horizontal'])
elif ymax - ymin == 1:
geometry = AcquisitionGeometry('parallel',
'2D',
angles,
pixel_num_h=dims[2],
pixel_size_h=1,
dist_source_center=None,
dist_center_detector=None,
channels=1)
return AcquisitionData(data.squeeze(),
False,
geometry=geometry,
dimension_labels=['angle', 'horizontal'])
plt.show()
ig = ImageGeometry(voxel_num_x = N, voxel_num_y = N, channels = np.shape(phantom_2Dt)[0])
data = ImageData(phantom_2Dt, geometry=ig)
detectors = N
angles = np.linspace(0,np.pi,180)
ag = AcquisitionGeometry('parallel','2D', angles, detectors, channels = np.shape(phantom_2Dt)[0])
Aop = AstraProjectorMC(ig, ag, dev)
sin = Aop.direct(data)
scale = 2
n1 = scale * np.random.poisson(sin.as_array()/scale)
noisy_data = AcquisitionData(n1, ag)
# Regularisation Parameter
alpha = 10
# Create operators
#op1 = Gradient(ig)
op1 = Gradient(ig, correlation='SpaceChannels')
op2 = Aop
# Create BlockOperator
operator = BlockOperator(op1, op2, shape=(2,1) )
# Create functions
f1 = alpha * MixedL21Norm()
ag = AcquisitionGeometry(
'cone',
'3D',
angles,
pixel_num_h=num_pixels_h,
pixel_size_h=0.25,
pixel_num_v=num_pixels_v,
pixel_size_v=0.25,
dist_source_center=233.0,
dist_center_detector=245.0,
channels=num_channels,
dimension_labels=['channel', 'vertical', 'angle', 'horizontal'])
# AcquisitionData 3D + channels
data = AcquisitionData(
X, ag, dimension_labels=['channel', 'vertical', 'angle', 'horizontal'])
# First need the geometric magnification to scale the voxel size relative
# to the detector pixel size.
# magnification factor = 250/1.76
# dist_source , dist center do not matter
ig = ImageGeometry(voxel_num_x=ag.pixel_num_h,
voxel_num_y=ag.pixel_num_h,
voxel_num_z=ag.pixel_num_h,
voxel_size_x=0.142,
voxel_size_y=0.142,
voxel_size_z=0.142,
channels=num_channels)
# Setup Astra Projector for 3DMC
# Set angles to use
angles = numpy.linspace(-numpy.pi, numpy.pi, num_angles, endpoint=False)
# Define full 3D acquisition geometry and data container.
# Geometric info is taken from the txt-file in the same dir as the mat-file
ag = AcquisitionGeometry('cone',
'3D',
angles,
pixel_num_h=num_pixels_h,
pixel_size_h=0.25,
pixel_num_v=num_pixels_v,
pixel_size_v=0.25,
dist_source_center=233.0,
dist_center_detector=245.0,
channels=num_channels)
data = AcquisitionData(X, geometry=ag)
# Reduce to central slice by extracting relevant parameters from data and its
# geometry. Perhaps create function to extract central slice automatically?
data2d = data.subset(vertical=40)
ag2d = AcquisitionGeometry('cone',
'2D',
ag.angles,
pixel_num_h=ag.pixel_num_h,
pixel_size_h=ag.pixel_size_h,
pixel_num_v=1,
pixel_size_v=ag.pixel_size_h,
dist_source_center=ag.dist_source_center,
dist_center_detector=ag.dist_center_detector,
channels=ag.channels)
data2d.geometry = ag2d
def process(self, out=None):
projections = self.get_input()
w = projections.get_dimension_size('horizontal')
delta = w - 2 * self.center_of_rotation
padded_width = int(numpy.ceil(abs(delta)) + w)
delta_pix = padded_width - w
voxel_per_pixel = 1
geom = pbalg.pb_setup_geometry_from_acquisition(
projections.as_array(), self.acquisition_geometry.angles,
self.center_of_rotation, voxel_per_pixel)
padded_geometry = self.acquisition_geometry.clone()
padded_geometry.pixel_num_h = geom['n_h']
padded_geometry.pixel_num_v = geom['n_v']
delta_pix_h = padded_geometry.pixel_num_h - self.acquisition_geometry.pixel_num_h
delta_pix_v = padded_geometry.pixel_num_v - self.acquisition_geometry.pixel_num_v
if delta_pix_h == 0:
delta_pix_h = delta_pix
padded_geometry.pixel_num_h = padded_width
#initialize a new AcquisitionData with values close to 0
out = AcquisitionData(geometry=padded_geometry)
out = out + self.pad_value
#pad in the horizontal-vertical plane -> slice on angles
if delta > 0:
#pad left of middle
command = "out.array["
for i in range(out.number_of_dimensions):
if out.dimension_labels[i] == 'horizontal':
value = '{0}:{1}'.format(delta_pix_h, delta_pix_h + w)
command = command + str(value)
else:
if out.dimension_labels[i] == 'vertical':
value = '{0}:'.format(delta_pix_v)
command = command + str(value)
else:
command = command + ":"
if i < out.number_of_dimensions - 1:
command = command + ','
command = command + '] = projections.array'
#print (command)
else:
#pad right of middle
command = "out.array["
for i in range(out.number_of_dimensions):
if out.dimension_labels[i] == 'horizontal':
value = '{0}:{1}'.format(0, w)
command = command + str(value)
else:
if out.dimension_labels[i] == 'vertical':
value = '{0}:'.format(delta_pix_v)
command = command + str(value)
else:
command = command + ":"
if i < out.number_of_dimensions - 1:
command = command + ','
command = command + '] = projections.array'
#print (command)
#cleaned = eval(command)
exec(command)
return out
if device == '1':
dev = 'gpu'
else:
dev = 'cpu'
Aop = AstraProjectorSimple(ig, ag, dev)
sin = Aop.direct(data)
# Create noisy data. Apply Gaussian noise
noises = ['gaussian', 'poisson']
noise = noises[which_noise]
if noise == 'poisson':
scale = 5
eta = 0
noisy_data = AcquisitionData(
np.random.poisson(scale * (eta + sin.as_array())) / scale, ag)
elif noise == 'gaussian':
n1 = np.random.normal(0, 1, size=ag.shape)
noisy_data = AcquisitionData(n1 + sin.as_array(), ag)
else:
raise ValueError('Unsupported Noise ', noise)
# Show Ground Truth and Noisy Data
plt.figure(figsize=(10, 10))
plt.subplot(1, 2, 2)
plt.imshow(data.as_array())
plt.title('Ground Truth')
plt.colorbar()
plt.subplot(1, 2, 1)
plt.imshow(noisy_data.as_array())
plt.title('Noisy Data')
# Apply centering correction by zero padding, amount found manually
cor_pad = 30
sino_pad = np.zeros((sino.shape[0],sino.shape[1]+cor_pad))
sino_pad[:,cor_pad:] = sino
# Extract AcquisitionGeometry for central slice for 2D fanbeam reconstruction
ag2d = AcquisitionGeometry('cone',
'2D',
angles=-np.pi/180*data.geometry.angles,
pixel_num_h=data.geometry.pixel_num_h + cor_pad,
pixel_size_h=data.geometry.pixel_size_h,
dist_source_center=-data.geometry.dist_source_center,
dist_center_detector=data.geometry.dist_center_detector)
# Set up AcquisitionData object for central slice 2D fanbeam
data2d = AcquisitionData(sino_pad,geometry=ag2d)
# Choose the number of voxels to reconstruct onto as number of detector pixels
N = data.geometry.pixel_num_h
# Geometric magnification
mag = (np.abs(data.geometry.dist_center_detector) + \
np.abs(data.geometry.dist_source_center)) / \
np.abs(data.geometry.dist_source_center)
# Voxel size is detector pixel size divided by mag
voxel_size_h = data.geometry.pixel_size_h / mag
# Construct the appropriate ImageGeometry
ig2d = ImageGeometry(voxel_num_x=N,
voxel_num_y=N,
path = os.path.dirname(tomophantom.__file__)
path_library2D = os.path.join(path, "Phantom2DLibrary.dat")
phantom_2D = TomoP2D.Model(model, N, path_library2D)
ig = ImageGeometry(voxel_num_x=N, voxel_num_y=N)
data = ImageData(phantom_2D)
detectors = N
angles = np.linspace(0, np.pi, 128, dtype=np.float32)
ag = AcquisitionGeometry('parallel', '2D', angles, detectors)
Aop = AstraProjectorSimple(ig, ag, dev)
sin = Aop.direct(data)
#noisy_data = AcquisitionData( sin.as_array() + np.random.normal(0,1,ag.shape))
noisy_data = AcquisitionData(sin.as_array())
# Show Ground Truth and Noisy Data
plt.figure(figsize=(10, 10))
plt.subplot(2, 1, 1)
plt.imshow(data.as_array())
plt.title('Ground Truth')
plt.colorbar()
plt.subplot(2, 1, 2)
plt.imshow(noisy_data.as_array())
plt.title('Noisy Data')
plt.colorbar()
plt.show()
# Setup and run the CGLS algorithm
alpha = 2
from ccpi.io.reader import XTEKReader
import numpy as np
import matplotlib.pyplot as plt
from ccpi.framework import ImageGeometry, AcquisitionData, ImageData
from ccpi.astra.ops import AstraProjector3DSimple
from ccpi.optimisation.algs import CGLS
import numpy
# Set up reader object and read the data
datareader = XTEKReader("REPLACE_THIS_BY_PATH_TO_DATASET/SophiaBeads_256_averaged.xtekct")
data = datareader.get_acquisition_data()
# Crop data and fix dimension labels
data = AcquisitionData(data.array[:,:,901:1101],
geometry=data.geometry,
dimension_labels=['angle','horizontal','vertical'])
data.geometry.pixel_num_v = 200
# Scale and negative-log transform
data.array = -np.log(data.as_array()/60000.0)
# Apply centering correction by zero padding, amount found manually
cor_pad = 30
data_pad = np.zeros((data.shape[0],data.shape[1]+cor_pad,data.shape[2]))
data_pad[:,cor_pad:,:] = data.array
data.geometry.pixel_num_h = data.geometry.pixel_num_h + cor_pad
data.array = data_pad
# Choose the number of voxels to reconstruct onto as number of detector pixels
N = data.geometry.pixel_num_h
angles = np.linspace(0, np.pi, 180, dtype=np.float32)
ag = AcquisitionGeometry('parallel', '2D', angles, detectors)
device = input('Available device: GPU==1 / CPU==0 ')
if device == '1':
dev = 'gpu'
else:
dev = 'cpu'
Aop = AstraProjectorSimple(ig, ag, dev)
sin = Aop.direct(data)
np.random.seed(10)
noisy_data = AcquisitionData(sin.as_array() + np.random.normal(0, 1, ag.shape))
# Show Ground Truth and Noisy Data
plt.figure(figsize=(10, 10))
plt.subplot(2, 1, 1)
plt.imshow(data.as_array())
plt.title('Ground Truth')
plt.colorbar()
plt.subplot(2, 1, 2)
plt.imshow(noisy_data.as_array())
plt.title('Noisy Data')
plt.colorbar()
plt.show()
# Setup and run the simple CGLS algorithm
x_init = ig.allocate()
def load_data(self):
'''
Parse NEXUS file and returns either ImageData or Acquisition Data
depending on file content
'''
try:
with h5py.File(self.nexus_file, 'r') as file:
if np.string_(file.attrs['creator']) != np.string_(
'NEXUSDataWriter.py'):
raise Exception(
'We can parse only files created by NEXUSDataWriter.py'
)
ds_data = file['entry1/tomo_entry/data/data']
data = np.array(ds_data, dtype='float32')
dimension_labels = []
for i in range(data.ndim):
dimension_labels.append(ds_data.attrs['dim{}'.format(i)])
if ds_data.attrs['data_type'] == 'ImageData':
self._geometry = ImageGeometry(
voxel_num_x=ds_data.attrs['voxel_num_x'],
voxel_num_y=ds_data.attrs['voxel_num_y'],
voxel_num_z=ds_data.attrs['voxel_num_z'],
voxel_size_x=ds_data.attrs['voxel_size_x'],
voxel_size_y=ds_data.attrs['voxel_size_y'],
voxel_size_z=ds_data.attrs['voxel_size_z'],
center_x=ds_data.attrs['center_x'],
center_y=ds_data.attrs['center_y'],
center_z=ds_data.attrs['center_z'],
channels=ds_data.attrs['channels'])
return ImageData(array=data,
deep_copy=False,
geometry=self._geometry,
dimension_labels=dimension_labels)
else: # AcquisitionData
if ds_data.attrs.__contains__('dist_source_center'):
dist_source_center = ds_data.attrs[
'dist_source_center']
else:
dist_source_center = None
if ds_data.attrs.__contains__('dist_center_detector'):
dist_center_detector = ds_data.attrs[
'dist_center_detector']
else:
dist_center_detector = None
self._geometry = AcquisitionGeometry(
geom_type=ds_data.attrs['geom_type'],
dimension=ds_data.attrs['dimension'],
dist_source_center=dist_source_center,
dist_center_detector=dist_center_detector,
pixel_num_h=ds_data.attrs['pixel_num_h'],
pixel_size_h=ds_data.attrs['pixel_size_h'],
pixel_num_v=ds_data.attrs['pixel_num_v'],
pixel_size_v=ds_data.attrs['pixel_size_v'],
channels=ds_data.attrs['channels'],
angles=np.array(
file['entry1/tomo_entry/data/rotation_angle'],
dtype='float32'))
#angle_unit = file['entry1/tomo_entry/data/rotation_angle'].attrs['units'])
return AcquisitionData(array=data,
deep_copy=False,
geometry=self._geometry,
dimension_labels=dimension_labels)
except:
print("Error reading nexus file")
raise
plt.imshow(b.array)
plt.title('Simulated data')
plt.colorbar()
plt.show()
plt.imshow(z.array)
plt.title('Backprojected data')
plt.colorbar()
plt.show()
One = ImageData(geometry=ig)
xOne = One.as_array()
xOne[:,:] = 1.0
OneD = AcquisitionData(geometry=ag)
y1 = OneD.as_array()
y1[:,:] = 1.0
s1 = (OneD*(Aop.direct(One))).sum()
s2 = (One*(Aop.adjoint(OneD))).sum()
print(s1)
print(s2)
print(s2/s1)
print((b*b).sum())
print((z*Phantom).sum())
print((z*Phantom).sum() / (b*b).sum())
#print(N/det_num)
#print(0.5*det_w/dx)
def SIRT(x_init, operator, data, opt=None, constraint=None):
'''Simultaneous Iterative Reconstruction Technique
Parameters:
x_init: initial guess
operator: operator for forward/backward projections
data: data to operate on
opt: additional algorithm
constraint: func of Indicator type specifying convex constraint.
'''
if opt is None:
opt = {'tol': 1e-4, 'iter': 1000}
else:
try:
max_iter = opt['iter']
except KeyError as ke:
opt[ke] = 1000
try:
opt['tol'] = 1000
except KeyError as ke:
opt[ke] = 1e-4
tol = opt['tol']
max_iter = opt['iter']
# Set default constraint to unconstrained
if constraint == None:
constraint = Function()
x = x_init.clone()
timing = numpy.zeros(max_iter)
criter = numpy.zeros(max_iter)
# Relaxation parameter must be strictly between 0 and 2. For now fix at 1.0
relax_par = 1.0
# Set up scaling matrices D and M.
im1 = ImageData(geometry=x_init.geometry)
im1.array[:] = 1.0
M = 1 / operator.direct(im1)
del im1
aq1 = AcquisitionData(geometry=M.geometry)
aq1.array[:] = 1.0
D = 1 / operator.adjoint(aq1)
del aq1
# algorithm loop
for it in range(0, max_iter):
t = time.time()
r = data - operator.direct(x)
x = constraint.prox(x + relax_par * (D * operator.adjoint(M * r)),
None)
timing[it] = time.time() - t
if it > 0:
criter[it - 1] = (r**2).sum()
r = data - operator.direct(x)
criter[it] = (r**2).sum()
return x, it, timing, criter
OrigDetec = 0
angles = np.linspace(0, 2 * np.pi, angles_num)
ag = AcquisitionGeometry('cone',
'2D',
angles,
detectors,
det_w,
dist_source_center=SourceOrig,
dist_center_detector=OrigDetec)
else:
NotImplemented
Aop = AstraProjectorSimple(ig, ag, dev)
sin = Aop.direct(data)
eta = 0
noisy_data = AcquisitionData(sin.as_array() + np.random.normal(0, 1, ag.shape))
back_proj = Aop.adjoint(noisy_data)
# Define Least Squares
f = FunctionOperatorComposition(L2NormSquared(b=noisy_data), Aop)
# Allocate solution
x_init = ig.allocate()
# Run FISTA for least squares
fista = FISTA(x_init=x_init, f=f, g=ZeroFunction())
fista.max_iteration = 10
fista.update_objective_interval = 2
fista.run(100, verbose=True)
# Run FISTA for least squares with lower/upper bound
data_rel = -numpy.log(data)
# Permute order to get: angles, vertical, horizontal, as default in framework.
data_rel = numpy.transpose(data_rel,(0,2,1))
# Set angles to use
angles = numpy.linspace(-numpy.pi,numpy.pi,num_angles,endpoint=False)
# Create 3D acquisition geometry and acquisition data
ag = AcquisitionGeometry('parallel',
'3D',
angles,
pixel_num_h=imsize,
pixel_num_v=imsize)
b = AcquisitionData(data_rel, geometry=ag)
# Reduce to single (noncentral) slice by extracting relevant parameters from data and its
# geometry. Perhaps create function to extract central slice automatically?
b2d = b.subset(vertical=128)
ag2d = AcquisitionGeometry('parallel',
'2D',
ag.angles,
pixel_num_h=ag.pixel_num_h)
b2d.geometry = ag2d
# Create 2D image geometry
ig2d = ImageGeometry(voxel_num_x=ag2d.pixel_num_h,
voxel_num_y=ag2d.pixel_num_h)
# Create GPU projector/backprojector operator with ASTRA.
ig = data.geometry
detectors = N
angles = np.linspace(0, np.pi, N, dtype=np.float32)
ag = AcquisitionGeometry('parallel', '2D', angles, detectors)
device = input('Available device: GPU==1 / CPU==0 ')
if device == '1':
dev = 'gpu'
else:
dev = 'cpu'
Aop = AstraProjectorSimple(ig, ag, dev)
sin = Aop.direct(data)
noisy_data = AcquisitionData(sin.as_array() + np.random.normal(0, 3, ig.shape))
# Setup and run the CGLS algorithm
alpha = 50
Grad = Gradient(ig)
# Form Tikhonov as a Block CGLS structure
op_CGLS = BlockOperator(Aop, alpha * Grad, shape=(2, 1))
block_data = BlockDataContainer(noisy_data, Grad.range_geometry().allocate())
x_init = ig.allocate()
cgls = CGLS(x_init=x_init, operator=op_CGLS, data=block_data)
cgls.max_iteration = 1000
cgls.update_objective_interval = 200
cgls.run(1000, verbose=False)
# Create Acquisition data
detectors = N
angles = np.linspace(0, np.pi, 180, dtype=np.float32)
ag = AcquisitionGeometry('parallel', '2D', angles, detectors)
device = input('Available device: GPU==1 / CPU==0 ')
if device == '1':
dev = 'gpu'
else:
dev = 'cpu'
Aop = AstraProjectorSimple(ig, ag, dev)
sin = Aop.direct(data)
projection_data = AcquisitionData(sin.as_array())
# Show Ground Truth and Noisy Data
plt.figure(figsize=(10, 10))
plt.subplot(2, 1, 1)
plt.imshow(data.as_array())
plt.title('Ground Truth')
plt.colorbar()
plt.subplot(2, 1, 2)
plt.imshow(projection_data.as_array())
plt.title('Projection Data')
plt.colorbar()
plt.show()
# Setup and run the CGLS algorithm
x_init = ig.allocate()
