-
Notifications
You must be signed in to change notification settings - Fork 0
/
ct_affreg_multigrid.py
309 lines (242 loc) · 13.3 KB
/
ct_affreg_multigrid.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
import numpy as np
import odl
import functionals as fctls
import algorithms as algs
import operators as ops
import deform as defm
import matplotlib.pyplot as plt
import skimage.transform as sktransform
from skimage.metrics import structural_similarity as ssim
import os
#%%
# Model parameters
stddev = 0.0015 # standard deviation for gaussian noise
num_angles = 200 # number of Radon projection angles
phis = [0.1] # rotation angle between side info and data
shift = [0.02, 0.08] # shift betwwen side info and data
noiseModel = 'poisson' # 'gaussian', 's-p', 'poisson'
noise = True
# Algorithmic parameters
save2Disk = True
niter = [10, 20, 50, 100, 200, 500] # number of PALM iterations per resolution
resolutions = [9,18, 36, 72, 144]
factor = 2
data_aligned = False
tol = 1e-9
impl = 'astra_cpu'
# Vectorfield parameters
gamma = 0.9995
eta = 1e-2
# Define reconstruction settings
settings = ['tv_recon', 'dtv_affreg_recon']
data_folder = '../data'
data_fnames = ['data1']
output_folder = '../results'
for data_fname in data_fnames:
for phi in phis:
for setting in settings:
if setting == 'tv_recon':
use_sinfo = False
ud_vars = [0]
alpha = 4e-2
elif setting == 'dtv_affreg_recon':
use_sinfo = True
ud_vars = [0,1]
alpha = 4e-1
else:
raise ValueError('Unknown setting')
folder_out = '{}/{}'.format(output_folder, data_fname)
folder_out += '_angles_{}'.format(num_angles)
folder_out += '_rot_{}'.format(phi)
folder_out += '_noiseModel_{}'.format(noiseModel)
folder_out += '_scale_space_{}'.format(len(resolutions))
folder_out += '_multigrid'
if not os.path.exists(folder_out):
os.makedirs(folder_out)
# Target space
gt, sinfo = np.load('{}/{}.npy'.format(data_folder, data_fname))
gt = np.pad(gt, ((12, 12), (12, 12)), 'constant',
constant_values=0.0014067312735875006)
sinfo = np.pad(sinfo, ((12, 12), (12, 12)), 'constant',
constant_values=0.02259887079937919)
simage = sinfo.shape
X = odl.uniform_discr([-1, -1], [1, 1], simage, interp='linear')
V = X.tangent_bundle
Yaff = odl.tensor_space(6)
Z = odl.ProductSpace(X, Yaff)
gt = X.element(gt)
sinfo = X.element(sinfo)
# Generate affine map and data image
cosp = np.cos(phi)
sinp = np.sin(phi)
disp_func = [
lambda x: (cosp-1)*x[0] - sinp*x[1] + shift[0],
lambda x: sinp*x[0] + (cosp-1)*x[1] + shift[1]]
vf_gt = Yaff.element([shift[0], shift[1], cosp-1, -sinp, sinp, cosp-1])
vf = V.element(disp_func)
X_plot = odl.uniform_discr([-1, -1], [1, 1], (15, 15), interp='linear')
V_plot = X_plot.tangent_bundle
x,y = X_plot.meshgrid
u = disp_func[0]([x,y])
v = disp_func[1]([x,y])
plt.quiver(x,y,u,v,color='yellow')
plt.axis('equal'),
plt.axis([-1.3,1.3,-1.3,1.3]),
plt.gca().set_aspect('equal', adjustable='box'),
plt.savefig(folder_out + '/deformation_field.png',
bbox_inches='tight',
pad_inches=0,
transparent=True, dpi=600,format='png')
deform_op = defm.LinDeformFixedTempl(gt)
# Create data
angle_partition = odl.uniform_partition(0, np.pi, num_angles)
detector_partition = odl.uniform_partition(-1.5, 1.5, int(1.6 * sinfo.shape[0]))
geometry = odl.tomo.Parallel2dGeometry(angle_partition, detector_partition)
A = odl.tomo.RayTransform(X, geometry, impl=impl)
FBP = odl.tomo.fbp_op(A, filter_type='Hann')
deformed_gt = deform_op(vf)
deformed_sino = A(deformed_gt)
aligned_sino = A(gt)
poisson_factor = 100/np.max(aligned_sino)
aligned_data = aligned_sino
deformed_data = deformed_sino
if noise:
if noiseModel == 'gaussian':
noise = odl.phantom.white_noise(A.range, mean=0, stddev=stddev, seed=1)
aligned_data += noise
deformed_data += noise
elif noiseModel == 's-p':
aligned_data = odl.phantom.salt_pepper_noise(aligned_data, seed=1)
deformed_data = odl.phantom.salt_pepper_noise(deformed_data, seed=1)
elif noiseModel == 'poisson':
background = 7 * aligned_sino.space.one()
aligned_data = odl.phantom.poisson_noise(poisson_factor*aligned_sino + background, seed=1807)
deformed_data = odl.phantom.poisson_noise(poisson_factor*deformed_sino + background, seed=1807)
else:
raise NotImplementedError('Unknown or missing noise type')
else:
if noiseModel == 'poisson':
background = 7 * aligned_sino.space.one()
aligned_data = poisson_factor * aligned_sino + background
deformed_data = poisson_factor * deformed_sino + background
file1 = open("{}.txt".format(folder_out + '/' + 'FBP'),"w")
if noiseModel == 'poisson':
ssim_val_fbp = ssim((FBP(aligned_data)),gt)
else:
ssim_val_fbp = ssim(FBP(aligned_data),gt)
file1.write("SSIM: {} \n".format(ssim_val_fbp))
file1.close()
# Save data images
plt.imsave(folder_out + '/' + data_fname + '_data.png', aligned_data, cmap='gray')
plt.imsave(folder_out + '/' + data_fname + '_data_deformed.png', deformed_data, cmap='gray')
plt.imsave(folder_out + '/' + data_fname + '_gt_deformed.png', deformed_gt, cmap='inferno')
plt.imsave(folder_out + '/' + data_fname + '_gt.png', gt, cmap='inferno')
plt.imsave(folder_out + '/' + data_fname + '_fbp_deformed.png', FBP(deformed_data), cmap='inferno')
plt.imsave(folder_out + '/' + data_fname + '_fbp.png', FBP(aligned_data), cmap='inferno')
plt.imsave(folder_out + '/' + data_fname + '_sinfo.png', sinfo, cmap='gray')
#%% reconstruction
X_first = odl.uniform_discr([-1, -1], [1, 1], (resolutions[0],resolutions[0]), interp='linear')
im_init = X_first.one()*1
vf_init = Yaff.zero()
ind = 0
basis = 10
alphas = [alpha*basis**i for i in range(len(resolutions))]
alphas.reverse()
for res in resolutions:
RegParam = alphas[ind]
simage_res = (res,res)
if use_sinfo is True:
sinfo_res = sktransform.resize(sinfo, simage_res)
else:
sinfo_res = None
X_res = odl.uniform_discr([-1, -1], [1, 1], simage_res, interp='linear')
V_res = X_res.tangent_bundle
Z_res = odl.ProductSpace(X_res, Yaff)
A_res = odl.tomo.RayTransform(X_res, geometry, impl=impl)
if noiseModel == 'poisson':
A_res *= poisson_factor
# Set some parameters and the general TV prox options
prox_options = {}
prox_options['name'] = 'FGP'
prox_options['warmstart'] = True
prox_options['p'] = None
prox_options['tol'] = None
prox_options['niter'] = 10 #100
strong_convexity = 0
# Define non-smooth part
reg_affine = odl.solvers.ZeroFunctional(Yaff)
# reg_affine = odl.solvers.IndicatorBox(Yaff,-0.1,0.1)
step = 10
if data_aligned is True:
data = aligned_data
else:
data = deformed_data
if noiseModel == 'gaussian':
datafit = 0.5 * odl.solvers.functional.L2NormSquared(data.space).translated(data)
elif noiseModel == 's-p':
datafit = odl.solvers.functional.LpNorm(data.space, 1).translated(data)
elif noiseModel == 'poisson':
datafit = odl.solvers.functional.KullbackLeibler(data.space, prior=data).translated(-background)
else:
raise NotImplementedError('Unkown noise type')
f = fctls.DataFitDisp(Z_res, datafit, forward=A_res)
file_out = '{}_da{}_si{}_vars{}_a{:4.2e}_g{:6.4f}_e{:6.4f}_res{}'.format(data_fname,
data_aligned, use_sinfo, ud_vars, RegParam, gamma, eta, res)
reg_im = fctls.TV(X_res, alpha=RegParam, sinfo=sinfo_res, NonNeg=True,
prox_options=prox_options.copy(), gamma=gamma, eta=eta)
g = odl.solvers.SeparableSum(reg_im, reg_affine)
# Define objective functional
obj = f + g
x_init = Z_res.element([im_init,vf_init])
cb = (odl.solvers.CallbackPrintIteration(end=', ') &
odl.solvers.CallbackPrintTiming(cumulative=False, end=', ') &
odl.solvers.CallbackPrintTiming(fmt='total={:.3f} s', cumulative=True) &
odl.solvers.CallbackPrint(step=step, func=obj, fmt='obj={!r}') &
odl.solvers.CallbackShow(step=step))
recon = algs.PALM(f, g, ud_vars=ud_vars.copy(), x=x_init.copy(), niter=None,
callback=cb, L=[1e2, 1e4], tol=tol)
niter_diff = [niter[0]] + list(np.diff(niter))
clim = [0, 1]
proj = odl.solvers.IndicatorBox(X_res, lower=clim[0], upper=clim[1]).proximal(1)
for ni, nid in zip(niter, niter_diff):
recon.run(nid)
if 1 in ud_vars:
vf = ops.Embedding_Affine(Yaff, V_plot)(recon.x[1])
plt.quiver(x, y, vf[0], vf[1], color='red'),
plt.axis('equal'),
plt.axis([-1.3,1.3,-1.3,1.3]),
# plt.axis('off'),
plt.gca().set_aspect('equal', adjustable='box'),
if save2Disk is True:
file_out_ni = '{}_{:04d}'.format(file_out, ni)
if 1 in ud_vars:
file_out_def_ni = '{}_def_{:04d}'.format(file_out, ni)
plt.savefig(folder_out + '/' + file_out_def_ni + '.png',
bbox_inches='tight',
pad_inches=0,
transparent=True,
dpi=600, format='png')
plt.imsave(folder_out + '/' + file_out_ni + '.png',
proj(recon.x[0]), cmap='inferno')
np.save('{}.npy'.format(folder_out + '/' + file_out_ni), list(recon.x))
if save2Disk is True:
rel_diff_vf = (recon.x[1]-vf_gt).norm()/vf_gt.norm()
file1 = open("{}.txt".format(folder_out + '/' + file_out_ni + '_def'),"w")
file1.write("RelDiff VF: {} \n".format(rel_diff_vf))
file1.close()
X_new = odl.uniform_discr([-1, -1], [1, 1], (2*res,2*res), interp='linear')
S = ops.Subsampling(X_new, X_res, 0)
im_init = S.adjoint(recon.x[0]).copy()
vf_init = recon.x[1].copy()
ind += 1
if save2Disk is True:
im_resc = recon.x[0]
ssim_val = ssim(im_resc, gt)
rel_diff_vf = (recon.x[1]-vf_gt).norm()/vf_gt.norm()
file1 = open("{}.txt".format(folder_out + '/' + file_out_ni),"w")
file1.write("SSIM: {} \n".format(ssim_val))
if use_sinfo is False:
ssim_val_def = ssim( im_resc, deformed_gt)
file1.write("SSIM def: {} \n".format(ssim_val_def))
file1.write("RelDiff VF: {} \n".format(rel_diff_vf))
file1.close()