def register_astra_geometry(proj_fix, proj_mov, geom_fix, geom_mov, subsamp=1):
"""
Compute a rigid transformation that makes sure that two reconstruction volumes are alligned.
Args:
proj_fix : projection data of the fixed volume
proj_mov : projection data of the fixed volume
geom_fix : projection data of the fixed volume
geom_mov : projection data of the fixed volume
Returns:
geom : geometry for the second reconstruction volume
"""
print('Computing a rigid tranformation between two datasets.')
# Find maximum vol size:
sz = numpy.array([proj_fix.shape, proj_mov.shape]).max(0)
sz += 10 # for safety...
vol1 = numpy.zeros(sz, dtype='float32')
vol2 = numpy.zeros(sz, dtype='float32')
projector.settings.bounds = [0, 10]
projector.settings.subsets = 10
projector.settings['mode'] = 'sequential'
projector.FDK(proj_fix, vol1, geom_fix)
projector.SIRT(proj_fix, vol1, geom_fix, iterations=5)
projector.FDK(proj_mov, vol2, geom_mov)
projector.SIRT(proj_mov, vol2, geom_mov, iterations=5)
display.slice(vol1, title='Fixed volume preview')
display.slice(vol1, title='Moving volume preview')
# Find transformation between two volumes:
R, T = register_volumes(vol1,
vol2,
subsamp=subsamp,
use_moments=True,
use_CG=True)
return R, T
def _sample_FDK_(projections, geometry, sample):
'''
Compute a subsampled version of FDK
'''
geometry_ = geometry.copy()
projections_ = projections[::sample[0], ::sample[2], ::sample[2]]
# Apply subsampling to detector and volume:
vol_sample = [sample[0], sample[1], sample[2]]
det_sample = [sample[0], sample[2], sample[2]]
geometry_['vol_sample'] = vol_sample
geometry_['det_sample'] = det_sample
volume = projector.init_volume(projections_)
# Do FDK without progress_bar:
projector.settings.progress_bar = False
projector.FDK(projections_, volume, geometry_)
projector.settings.progress_bar = True
return volume
]
# Sinogram that will be simultated:
counts = numpy.zeros([1, 128, 128], dtype='float32')
# Simulate:
model.forward_spectral(vol, counts, geom, mats, E, S, n_phot=1e6)
proj = -numpy.log(counts)
# Display:
display.slice(proj, title='Modelled sinogram')
#%% Reconstruct:
vol_rec = numpy.zeros_like(vol)
projector.FDK(proj, vol_rec, geom)
display.slice(vol_rec, title='Uncorrected FDK')
display.plot(vol_rec[0, 64], title='Crossection')
#%% Estimate system spectrum:
print('Callibrating spectrum...')
energy, spec = analyze.calibrate_spectrum(proj,
vol,
geom,
compound='Al',
density=2.7)
#meta = {'energy':e, 'spectrum':s, 'description':geom.description}
#data.write_toml(os.path.join(path, 'spectrum.toml'), meta)
#%% Unfiltered back-project # Make volume: vol_rec = numpy.zeros_like(vol) # Backproject: projector.backproject(proj, vol_rec, geom) display.slice(vol_rec, title = 'Backprojection') #%% Filtered back-project # Make volume: vol_rec = numpy.zeros_like(vol) # Use FDK: projector.FDK(proj, vol_rec, geom) display.slice(vol_rec, title = 'FDK') #%% Simple SIRT: vol_rec = numpy.zeros_like(vol) projector.SIRT(proj, vol_rec, geom, iterations = 20) display.slice(vol_rec, title = 'SIRT') #%% SIRT with subsets and non-negativity: # Settings: projector.settings.update_residual = True projector.settings.bounds = [0, 2] projector.settings.subsets = 10 projector.settings.sorting = 'equidistant'
proj_a = numpy.zeros([128, 32, 128], dtype='float32')
proj_b = numpy.zeros([128, 32, 128], dtype='float32')
# Forward project:
projector.forwardproject(proj_a, vol, geom_a)
projector.forwardproject(proj_b, vol, geom_b)
display.slice(proj_a, dim=1, title='Proj A')
display.slice(proj_b, dim=1, title='Proj B')
#%% Preview reconstructions:
geom_b = geom_a.copy()
# First volume:
vola = projector.init_volume(proj_a)
projector.FDK(proj_a, vola, geom_a)
# Second volume:
volb = projector.init_volume(proj_b)
projector.FDK(proj_b, volb, geom_b)
display.projection(vola, dim=1, title='Volume A')
display.projection(volb, dim=1, title='Volume B')
#%% Register:
R, T = process.register_volumes(vola,
volb,
subsamp=1,
use_moments=True,
use_CG=True)
display.slice(vol, title='Phantom')
# Forward project:
projector.forwardproject(proj, vol, geom)
display.slice(proj, dim=1, title='Projection')
#%% Use optimize_rotation_center:
# Unmodified geometry:
geom = geometry.circular(src2obj=100,
det2obj=100,
det_pixel=0.1,
ang_range=[0, 360])
vol = projector.init_volume(proj)
projector.FDK(proj, vol, geom)
display.slice(vol, bounds=[0, 2], title='FDK: uncorrected')
#%% Optimization:
vals = numpy.linspace(0., 3., 7)
process.optimize_modifier(vals,
proj,
geom,
samp=[1, 1, 1],
key='det_roll',
metric='highpass')
#%% Reconstruct:
vol = projector.init_volume(proj)
phase_ctf *= model.ctf(proj.shape[::2], 'gaussian', [det_pixel, sigma * 1])
# Electro-magnetic field image:
proj_i = numpy.exp(-proj * n)
# Field intensity:
data.convolve_filter(proj_i, phase_ctf)
proj_i = numpy.abs(proj_i)**2
display.slice(proj_i, title='Sinogram (phase contrast)')
#%% Reconstruct directly:
vol_rec = numpy.zeros_like(vol)
projector.FDK(-numpy.log(proj_i), vol_rec, geom)
display.slice(vol_rec, title='FDK (Raw)')
#%% Invertion of phase contrast based on dual-CTF model:
# Propagator (Dual CTF):
alpha = numpy.imag(n) / numpy.real(n)
dual_ctf = model.ctf(proj.shape[::2], 'dual_ctf',
[det_pixel, energy, src2obj, det2obj, alpha])
dual_ctf *= model.ctf(proj.shape[::2], 'gaussian', [det_pixel, sigma])
# Use inverse convolution to solve for blurring and phase contrast
data.deconvolve_filter(proj_i, dual_ctf, epsilon=0.1)
# Depending on epsilon there is some lof frequency bias introduced...
proj_i /= proj_i.max()