parser = OptionParser(usage="%name -n NUMBER_OF_ITERATIONS -f DIFFRACTION_PATTERN -r REAL_SPACE_MODEL -s SUPPORT [-o OUTPUT_DIR -a AFFIX]")
parser.add_option("-n", action="store", type="int", default=100, dest="number_of_iterations", help="Number of iterations of ER")
parser.add_option("-f", action="store", type="string", default=None, dest="pattern", help="Diffraction pattern")
parser.add_option("-r", action="store", type="string", default=None, dest="real_space", help="Starting real-space model")
parser.add_option("-s", action="store", type="string", default=None, dest="support", help="Support")
parser.add_option("-o", action="store", type="string", default=".", dest="output_dir", help="Output directory")
parser.add_option("-a", action="store", type="string", default="refined", dest="output_affix", help="This name will be added to all output files.")
options, args = parser.parse_args()
_numpy.random.seed()
_spimage.sp_srand(_numpy.random.randint(1e6))
intensities = _spimage.sp_image_read(options.pattern, 0)
if intensities.shifted == 0:
amplitudes = _spimage.sp_image_shift(intensities)
else:
amplitudes = _spimage.sp_image_duplicate(intensities, _spimage.SP_COPY_ALL)
_spimage.sp_image_dephase(amplitudes)
_spimage.sp_image_to_amplitudes(amplitudes)
real_space = _spimage.sp_image_read(options.real_space, 0)
support = _spimage.sp_image_read(options.support, 0)
phase_alg = _spimage.sp_phasing_er_alloc(_spimage.SpNoConstraints)
sup_alg = _spimage.sp_support_array_init(_spimage.sp_support_static_alloc(), 20)
# create phaser
phaser = _spimage.sp_phaser_alloc()
_spimage.sp_phaser_init(phaser, phase_alg, sup_alg, _spimage.SpEngineCUDA)
_spimage.sp_phaser_set_amplitudes(phaser, amplitudes)
_spimage.sp_phaser_init_model(phaser, real_space, 0)
def prtf(images_rs,supports,translate=True,enantio=True,full_out=False):
"""
NOTE: For using the enantio option, the images need to be centered in fourier space (no phase ramp in real space)
"""
S = images_rs.shape
s = list(S)
N = s.pop(0)
s = tuple(s)
image0_rs = images_rs[0]
image0_fs = numpy.fft.fftn(image0_rs)
sp_image0_rs = numpy_array_to_image(image0_rs,supports[0])
sp_image0_fs = numpy_array_to_image(image0_fs)
sp_amp_fs = spimage.sp_image_duplicate(sp_image0_fs,spimage.SP_COPY_ALL)
spimage.sp_image_dephase(sp_amp_fs)
spimage.sp_image_free(sp_image0_rs)
spimage.sp_image_free(sp_image0_fs)
sum_fs = image0_fs.copy()
sum_fs[abs(sum_fs) > 0.] /= abs(sum_fs[abs(sum_fs) > 0.])
sp_sum_fs = numpy_array_to_image(sum_fs)
zeros = numpy.zeros(shape=s,dtype="int")
zeros[abs(sum_fs) <= 0.] = 1
sp_avg_img = numpy_array_to_image(image0_rs)
avg_msk = numpy.zeros(shape=s,dtype="float")
images_rs_super = numpy.zeros(shape=S,dtype="complex128")
images_rs_super[0,:] = image0_rs[:]
masks_rs_super = numpy.zeros(shape=S,dtype="bool")
masks_rs_super[0,:] = supports[0,:]
for i,img,sup in zip(range(1,N),images_rs[1:],supports[1:]):
# Initialize image
sp_img = numpy_array_to_image(img,sup)
# Translate and enantio matching
if translate:
spimage.sp_image_superimpose(sp_avg_img,sp_img, spimage.SpEnantiomorph if enantio else 0)
spimage.sp_image_phase_match(sp_avg_img,sp_img,2)
spimage.sp_image_add(sp_avg_img,sp_img)
if sp_img.mask.sum() > 0:
avg_msk[sp_img.mask == 0] += 1
# Cache image and support
images_rs_super[i,:] = sp_img.image[:]
masks_rs_super[i,:] = sp_img.mask[:]
# Add amplitudes
sp_tmp = spimage.sp_image_fftw3(sp_img)
sp_tmpamp = spimage.sp_image_duplicate(sp_tmp,spimage.SP_COPY_ALL)
spimage.sp_image_dephase(sp_tmpamp);
spimage.sp_image_add(sp_amp_fs,sp_tmpamp)
# Count zeros
positive = abs(sp_tmp.image) > 0.
sp_tmp.image[positive] /= abs(sp_tmp.image)[positive]
zeros += (positive == False)
spimage.sp_image_add(sp_sum_fs,sp_tmp)
spimage.sp_image_free(sp_img)
spimage.sp_image_free(sp_tmp)
spimage.sp_image_free(sp_tmpamp)
sp_prtf = spimage.sp_image_duplicate(sp_sum_fs,spimage.SP_COPY_DATA|spimage.SP_COPY_MASK)
sp_prtf.image[:] /= N
sp_prtf.image[zeros > 0] = 0.
spimage.sp_image_dephase(sp_prtf)
avg_img = sp_avg_img.image[:].copy()
avg_sup = avg_msk > 0
prtf = abs(sp_prtf.image[:]).copy()
prtf = numpy.fft.fftshift(prtf)
for sp_i in [sp_prtf,sp_avg_img,sp_amp_fs,sp_sum_fs]:
spimage.sp_image_free(sp_i)
out = {}
out["prtf"] = prtf
out["super_image"] = avg_img
if full_out:
out["prtf_r"] = spimage.radial_mean(prtf,cx=s[1]/2,cy=s[0]/2)
out["super_mask"] = avg_sup
out["images"] = images_rs_super
out["masks"] = masks_rs_super
return out
def prtf(images_rs, supports, translate=True, enantio=True, full_out=False):
"""
NOTE: For using the enantio option, the images need to be centered in fourier space (no phase ramp in real space)
"""
S = images_rs.shape
s = list(S)
N = s.pop(0)
s = tuple(s)
image0_rs = images_rs[0]
image0_fs = numpy.fft.fftn(image0_rs)
sp_image0_rs = numpy_array_to_image(image0_rs, supports[0])
sp_image0_fs = numpy_array_to_image(image0_fs)
sp_amp_fs = spimage.sp_image_duplicate(sp_image0_fs, spimage.SP_COPY_ALL)
spimage.sp_image_dephase(sp_amp_fs)
spimage.sp_image_free(sp_image0_rs)
spimage.sp_image_free(sp_image0_fs)
sum_fs = image0_fs.copy()
sum_fs[abs(sum_fs) > 0.] /= abs(sum_fs[abs(sum_fs) > 0.])
sp_sum_fs = numpy_array_to_image(sum_fs)
zeros = numpy.zeros(shape=s, dtype="int")
zeros[abs(sum_fs) <= 0.] = 1
sp_avg_img = numpy_array_to_image(image0_rs)
avg_msk = numpy.zeros(shape=s, dtype="float")
images_rs_super = numpy.zeros(shape=S, dtype="complex128")
images_rs_super[0, :] = image0_rs[:]
masks_rs_super = numpy.zeros(shape=S, dtype="bool")
masks_rs_super[0, :] = supports[0, :]
for i, img, sup in zip(range(1, N), images_rs[1:], supports[1:]):
# Initialize image
sp_img = numpy_array_to_image(img, sup)
# Translate and enantio matching
if translate:
spimage.sp_image_superimpose(
sp_avg_img, sp_img, spimage.SpEnantiomorph if enantio else 0)
spimage.sp_image_phase_match(sp_avg_img, sp_img, 2)
spimage.sp_image_add(sp_avg_img, sp_img)
if sp_img.mask.sum() > 0:
avg_msk[sp_img.mask == 0] += 1
# Cache image and support
images_rs_super[i, :] = sp_img.image[:]
masks_rs_super[i, :] = sp_img.mask[:]
# Add amplitudes
sp_tmp = spimage.sp_image_fftw3(sp_img)
sp_tmpamp = spimage.sp_image_duplicate(sp_tmp, spimage.SP_COPY_ALL)
spimage.sp_image_dephase(sp_tmpamp)
spimage.sp_image_add(sp_amp_fs, sp_tmpamp)
# Count zeros
positive = abs(sp_tmp.image) > 0.
sp_tmp.image[positive] /= abs(sp_tmp.image)[positive]
zeros += (positive == False)
spimage.sp_image_add(sp_sum_fs, sp_tmp)
spimage.sp_image_free(sp_img)
spimage.sp_image_free(sp_tmp)
spimage.sp_image_free(sp_tmpamp)
sp_prtf = spimage.sp_image_duplicate(
sp_sum_fs, spimage.SP_COPY_DATA | spimage.SP_COPY_MASK)
sp_prtf.image[:] /= N
sp_prtf.image[zeros > 0] = 0.
spimage.sp_image_dephase(sp_prtf)
avg_img = sp_avg_img.image[:].copy()
avg_sup = avg_msk > 0
prtf = abs(sp_prtf.image[:]).copy()
prtf = numpy.fft.fftshift(prtf)
for sp_i in [sp_prtf, sp_avg_img, sp_amp_fs, sp_sum_fs]:
spimage.sp_image_free(sp_i)
out = {}
out["prtf"] = prtf
out["super_image"] = avg_img
if full_out:
out["prtf_r"] = spimage.radial_mean(prtf, cx=s[1] / 2, cy=s[0] / 2)
out["super_mask"] = avg_sup
out["images"] = images_rs_super
out["masks"] = masks_rs_super
return out
default=100,
help="Number of iterations of ER")
parser.add_argument("-o", "--outdir", default=".", help="Output directory")
parser.add_argument("-a",
"--affix",
default="refined",
help="This name will be added to all output files.")
args = parser.parse_args()
_numpy.random.seed()
_spimage.sp_srand(_numpy.random.randint(1e6))
intensities = _spimage.sp_image_read(args.pattern, 0)
if intensities.shifted == 0:
amplitudes = _spimage.sp_image_shift(intensities)
else:
amplitudes = _spimage.sp_image_duplicate(intensities,
_spimage.SP_COPY_ALL)
_spimage.sp_image_dephase(amplitudes)
_spimage.sp_image_to_amplitudes(amplitudes)
real_space = _spimage.sp_image_read(args.real_space, 0)
support = _spimage.sp_image_read(args.support, 0)
phase_alg = _spimage.sp_phasing_er_alloc(_spimage.SpNoConstraints)
sup_alg = _spimage.sp_support_array_init(
_spimage.sp_support_static_alloc(), 20)
# create phaser
phaser = _spimage.sp_phaser_alloc()
_spimage.sp_phaser_init(phaser, phase_alg, sup_alg, _spimage.SpEngineCUDA)
_spimage.sp_phaser_set_amplitudes(phaser, amplitudes)
