vol_geom = astra.create_vol_geom(256, 256)
proj_geom = astra.create_proj_geom('parallel', 1.0, 256,
np.linspace(0, np.pi, 32, False))
# Create the ASTRA projector (change 'linear' to 'cuda' to use GPU)
pid = astra.create_projector('linear', proj_geom, vol_geom)
p = astra.OpTomo(pid)
# Load the phantom from disk
testPhantom = np.load('phantom.npy')
# Calculate the forward projection of the phantom
testSino = (p * testPhantom).reshape(p.sshape)
# Add some noise to the sinogram
testSino = astra.add_noise_to_sino(testSino, 10**4)
# Reconstruct the image using MR-FBP, FBP, and SIRT.
mrRec = p.reconstruct('MR-FBP', testSino)
if astra.projector.is_cuda(pid):
fbpRec = p.reconstruct('FBP_CUDA', testSino)
sirtRec = p.reconstruct('SIRT_CUDA', testSino, 200)
else:
fbpRec = p.reconstruct('FBP', testSino)
sirtRec = p.reconstruct('SIRT', testSino, 200)
# Show the different reconstructions on screen
import pylab
pylab.gray()
pylab.subplot(221)
pylab.axis('off')
import astra
fl = open('example_phantom.json', 'r')
strg = fl.read()
fl.close()
oplist = json.loads(strg)
ang = np.linspace(0, np.pi, 256, False)
sz = 256
fp, imgs = lorentzphantom.generate.fp(oplist,
np.linspace(0, 1, len(ang)),
ang, (sz, sz),
fpfunc=lorentzphantom.forwproj.astrafp)
fp = astra.add_noise_to_sino(fp, 10**3)
proj_geom = astra.create_proj_geom('parallel', 1.0, sz, ang)
vol_geom = astra.create_vol_geom(sz)
pid = astra.create_projector('cuda', proj_geom, vol_geom)
w = astra.OpTomo(pid)
rec = w.reconstruct('FBP_CUDA', fp, iterations=1)
import pylab as pl
pl.gray()
pl.imshow(fp)
pl.show()
pl.imshow(rec)
pl.show()
na = 32
# Create ASTRA geometries
vol_geom = astra.create_vol_geom(nd,nd)
proj_geom = astra.create_proj_geom('parallel',1.0,nd,np.linspace(0,np.pi,na,False))
pi = astra.create_projector('linear', proj_geom, vol_geom) # change 'linear' to 'cuda' for GPU
p = astra.OpTomo(pi)
# Load the phantom from disk
testPhantom = phantom.phantom(nd)
# Calculate the forward projection of the phantom
testSino = (p*testPhantom).reshape((na,nd))
# Add some noise to the sinogram
testSino = astra.add_noise_to_sino(testSino,10**3)
# Register plugin with ASTRA
astra.plugin.register(sirtfbp.plugin)
# Reconstruct the image using SIRT-FBP, FBP, and SIRT.
sfbpRec = p.reconstruct('SIRT-FBP',testSino,200, extraOptions={'filter_dir':os.getcwd()})
if astra.projector.is_cuda(pi):
fbpRec = p.reconstruct('FBP_CUDA',testSino)
sirtRec = p.reconstruct('SIRT_CUDA',testSino,200)
else:
fbpRec = p.reconstruct('FBP',testSino)
sirtRec = p.reconstruct('SIRT',testSino,200)
# Show the different reconstructions on screen
import pylab
# Create the ASTRA projector
pid = astra.create_projector('linear',proj_geom,vol_geom) # change 'linear' to 'cuda' to use GPU
p = astra.OpTomo(pid)
# Create simulated HQ dataset for training
hq = np.zeros((ns,nd,nd))
hq[:,nd//4:3*nd//4,nd//4:3*nd//4]=1
hq[:,3*nd//8:5*nd//8,3*nd//8:5*nd//8]=0
try:
os.mkdir('hqrecs/')
except OSError:
pass
projections = np.zeros((ns, na, nd))
for i in range(ns):
projections[i] = astra.add_noise_to_sino((p*hq[i]).reshape(p.sshape),i0)
tifffile.imsave('hqrecs/{:04d}.tiff'.format(i),hq[i])
# Prepare training files
astra.plugin.register(nnfbp.plugin_prepare)
try:
os.mkdir('trainfiles/')
except OSError:
pass
for i in range(ns):
p.reconstruct('NN-FBP-prepare',projections[i],extraOptions={'hqrecfiles':'hqrecs/*.tiff','z_id':i,'traindir':'trainfiles/','npick':1000})
# Train filters and weights
nnfbp.plugin_train('trainfiles/', 4, 'filters.mat')
# Generate test sinogram
import lorentzphantom
import numpy as np
import json
import astra
fl = open('example_phantom.json','r')
strg = fl.read()
fl.close()
oplist = json.loads(strg)
ang = np.linspace(0,np.pi,256,False)
sz = 256
fp,imgs = lorentzphantom.generate.fp(oplist,np.linspace(0,1,len(ang)),ang,(sz,sz), fpfunc=lorentzphantom.forwproj.astrafp)
fp = astra.add_noise_to_sino(fp,10**3)
proj_geom = astra.create_proj_geom('parallel',1.0,sz,ang)
vol_geom = astra.create_vol_geom(sz)
pid = astra.create_projector('cuda',proj_geom,vol_geom)
w = astra.OpTomo(pid)
rec = w.reconstruct('FBP_CUDA',fp,iterations=1)
import pylab as pl
pl.gray()
pl.imshow(fp)
pl.show()
pl.imshow(rec)
pl.show()
import numpy as np
import astra
# Simple phantom
x = np.zeros((1024,1024))
x[400:600,400:600]=1
# ASTRA geometries, projector, and OpTomo
proj_geom = astra.create_proj_geom('parallel',1.0,1024,np.linspace(0,np.pi,180))
vol_geom = astra.create_vol_geom(1024)
pid = astra.create_projector('cuda',proj_geom,vol_geom)
p = astra.OpTomo(pid)
# Create sinogram and add noise
sino = (p*x).reshape(p.sshape)
sino_with_noise = astra.add_noise_to_sino(sino,10**2)
# Define parameters and FISTA object that performs reconstruction
lam = 2**-5
f = tvtomo.FISTA(p,lam,100)
# Actually reconstruct (takes time)
tvrec = f.reconstruct(sino_with_noise,50, progress=True)
# ASTRA data for reconstructions below
sid = astra.data2d.create('-sino', proj_geom,sino_with_noise)
rid = astra.data2d.create('-vol', vol_geom)
#ASTRA SIRT
cfg = astra.astra_dict('SIRT_CUDA')
cfg['ProjectionDataId'] = sid
import astra
# Simple phantom
x = np.zeros((1024, 1024))
x[400:600, 400:600] = 1
# ASTRA geometries, projector, and OpTomo
proj_geom = astra.create_proj_geom('parallel', 1.0, 1024,
np.linspace(0, np.pi, 180))
vol_geom = astra.create_vol_geom(1024)
pid = astra.create_projector('cuda', proj_geom, vol_geom)
p = astra.OpTomo(pid)
# Create sinogram and add noise
sino = (p * x).reshape(p.sshape)
sino_with_noise = astra.add_noise_to_sino(sino, 10**2)
# Define parameters and FISTA object that performs reconstruction
lam = 2**-5
f = tvtomo.FISTA(p, lam, 100)
# Actually reconstruct (takes time)
tvrec = f.reconstruct(sino_with_noise, 50, progress=True)
# ASTRA data for reconstructions below
sid = astra.data2d.create('-sino', proj_geom, sino_with_noise)
rid = astra.data2d.create('-vol', vol_geom)
#ASTRA SIRT
cfg = astra.astra_dict('SIRT_CUDA')
cfg['ProjectionDataId'] = sid
pid = astra.create_projector('linear', proj_geom,
vol_geom) # change 'linear' to 'cuda' to use GPU
p = astra.OpTomo(pid)
# Create simulated HQ dataset for training
hq = np.zeros((ns, nd, nd))
hq[:, nd // 4:3 * nd // 4, nd // 4:3 * nd // 4] = 1
hq[:, 3 * nd // 8:5 * nd // 8, 3 * nd // 8:5 * nd // 8] = 0
try:
os.mkdir('hqrecs/')
except OSError:
pass
projections = np.zeros((ns, na, nd))
for i in range(ns):
projections[i] = astra.add_noise_to_sino((p * hq[i]).reshape(p.sshape), i0)
tifffile.imsave('hqrecs/{:04d}.tiff'.format(i), hq[i])
# Prepare training files
astra.plugin.register(nnfbp.plugin_prepare)
try:
os.mkdir('trainfiles/')
except OSError:
pass
for i in range(ns):
p.reconstruct('NN-FBP-prepare',
projections[i],
extraOptions={
'hqrecfiles': 'hqrecs/*.tiff',
'z_id': i,
'traindir': 'trainfiles/',
