def unwrap_plots():
x, y = np.ogrid[:32, :32]
phi = 2*np.pi*(x*0.2 + y*0.1)
#phi = 1*np.arctan2(x-14.3, y-6.3) - 2*np.arctan2(x-18.3, y-22.1)
phi[8,8] = np.NaN
phi_wrapped = np.angle(np.exp(1j*phi))
phi_unwrapped = unwrap(phi_wrapped,
#wrap_around_axis_0 = True,
#wrap_around_axis_1 = True,
)
mask = np.zeros_like(phi, dtype = np.uint8)
#mask[10:22, 4:10] = 1
phi_wrapped_masked = np.ma.array(phi_wrapped, dtype = np.float32, mask = mask)
phi_unwrapped_masked = unwrap(phi_wrapped_masked)
import matplotlib.pyplot as plt
plt.figure(1)
plt.clf()
plt.gray()
plt.subplot(221)
plt.imshow(phi, interpolation = 'nearest')
plt.subplot(222)
plt.imshow(phi_wrapped, interpolation = 'nearest')
plt.subplot(223)
plt.imshow(phi_unwrapped, interpolation = 'nearest')
plt.subplot(224)
plt.imshow(phi_unwrapped_masked, interpolation = 'nearest')
plt.draw()
plt.show()
def calculate_phase_diff_map_1d(dY, dY0, th, ns, mask_for_unwrapping=None):
"""
TODO: Docstring for calculate_phase_diff_map_1d
# Basic FTP treatment.
# This function takes a deformed and a reference image and calculates the phase difference map between the two.
#
# INPUTS:
# dY = deformed image
# dY0 = reference image
# ns = size of gaussian filter
#
# OUTPUT:
# dphase = phase difference map between images
"""
ny, nx = np.shape(dY)
phase0 = np.zeros([nx, ny])
phase = np.zeros([nx, ny])
for lin in range(0, nx):
fY0 = np.fft.fft(dY0[lin, :])
fY = np.fft.fft(dY[lin, :])
dfy = 1. / ny
fy = np.arange(dfy, 1, dfy)
imax = np.argmax(np.abs(fY0[9:nx // 2]))
ifmax = imax + 9
HW = np.round(ifmax * th)
HW *= 0.5 # TODO
W = 2 * HW
win = signal.tukey(int(W), ns)
gaussfilt1D = np.zeros(nx)
gaussfilt1D[int(ifmax - HW - 1):int(ifmax - HW + W - 1)] = win
Nfy0 = fY0 * gaussfilt1D
Nfy = fY * gaussfilt1D
Ny0 = np.fft.ifft(Nfy0)
Ny = np.fft.ifft(Nfy)
phase0[lin, :] = np.angle(Ny0)
phase[lin, :] = np.angle(Ny)
if mask_for_unwrapping is None:
mphase0 = unwrap(phase0)
mphase = unwrap(phase)
else:
mphase0 = ma.masked_array(phase0, mask=mask_for_unwrapping)
mphase = ma.masked_array(phase, mask=mask_for_unwrapping)
mphase0 = unwrap(mphase0)
mphase = unwrap(mphase)
dphase = (mphase - mphase0)
return dphase
def rescale_and_unwrap_3D(img,roi_mask):
#First rescale:
img=(np.float32(img)-2048)*(np.pi/2048)
#Then unwrap:
import unwrap
img_unwrapped=unwrap.unwrap(np.ma.masked_array(img,roi_mask))
return img_unwrapped
def rescale_and_unwrap_3D(img, roi_mask):
# First rescale:
img = (np.float32(img) - 2048) * (np.pi / 2048)
# Then unwrap:
import unwrap
img_unwrapped = unwrap.unwrap(np.ma.masked_array(img, roi_mask))
return img_unwrapped
def pypi_unwrap(phase):
unwrapped_phase = unwrap.unwrap(phase,
wrap_around_axis_0=False,
wrap_around_axis_1=False,
wrap_around_axis_2=False)
return unwrapped_phase
def get_st_diagram(med_folder_name):
hdf5_path = f'../../../Mediciones/{med_folder_name}/HDF5/ST.hdf5'
f = h5py.File(hdf5_path, 'r')
st_diagram = np.array(f['spatiotemporal_diagram'])
gap = st_diagram[:, 100] - np.unwrap(st_diagram[:, 100])
gap = np.expand_dims(gap, 1)
st_diagram = st_diagram - gap
st_diagram = unwrap(st_diagram)
st_diagram -= np.expand_dims(st_diagram.mean(1), 1)
st_diagram[(st_diagram > 5) + (st_diagram < -5)] = np.nan
st_diagram = fill_nans(st_diagram)
st_diagram *= -1
# st_diagram = np.remainder(st_diagram, 2*np.pi)
# st_diagram[st_diagram < 1.2] = np.nan
# st_diagram = fill_nans(st_diagram)
# st_diagram = np.pi - st_diagram
# st_diagram = (st_diagram.T - st_diagram.mean(1)).T
# mask = (st_diagram > 5) + (st_diagram < -5)
# st_diagram[mask] = np.nan
# st_diagram = fill_nans(st_diagram)
# st_diagram *= -1
return st_diagram
def joint_likelihood_integrate(combined_params, Y, resp, example_params_struct, prior):
# First, unpack parameters.
cur_params = rewrap( example_params_struct, combined_params )
(nll_mix, dnll_mix_X) = mixture_likelihood_integrate(cur_params['X'], resp, prior)
##checkgrad('mixture_likelihood_integrate', cur_params.X, 1e-6, resp);
(nll_lvm, dnll_lvm_X, dnll_log_hypers) = gplvm_likelihood(cur_params.X, Y, cur_params['log_hypers'])
##checkgrad('gplvm_likelihood', combined_params, 1e-6);
##[ nll_back, dnll_back_X, dnll_log_hypers_back ] = ...
## back_constraint_likelihood( cur_params.X, Y, cur_params.log_hypers );
nll_back = 0
dnll_back_X = 0
##checkgrad('back_constraint_likelihood', cur_params.X, 1e-6, Y,
##cur_params.log_hypers);
nll = nll_lvm + nll_mix + nll_back
## Put gradients back into a vector.
all_grads_struct['X'] = np.matrix(dnll_lvm_X) + np.matrix(dnll_mix_X) + dnll_back_X
all_grads_struct['log_hypers'] = dnll_log_hypers
dnll = unwrap(all_grads_struct)
return (nll, dnll)
def test_unwrap2D(self):
x, y = np.ogrid[:8, :16]
phi = 2*np.pi*(x*0.2 + y*0.1)
phi_wrapped = np.angle(np.exp(1j*phi))
phi_unwrapped = unwrap(phi_wrapped)
s = np.round(phi_unwrapped[0,0]/(2*np.pi))
assert_array_almost_equal(phi, phi_unwrapped - s*2*np.pi)
def test_unwrap2D(self):
x, y = np.ogrid[:8, :16]
phi = 2 * np.pi * (x * 0.2 + y * 0.1)
phi_wrapped = np.angle(np.exp(1j * phi))
phi_unwrapped = unwrap(phi_wrapped)
s = np.round(phi_unwrapped[0, 0] / (2 * np.pi))
assert_array_almost_equal(phi, phi_unwrapped - s * 2 * np.pi)
def test_unwrap3D_y_wraparound(self):
"""
Regression test for an incorrect addressing when wrapping around Y-axis is enabled.
Produced a segfault.
"""
x, y, z = np.ogrid[:11, :128, :128]
phi = 2 * np.pi * (x * 0.2 + y * 0.1 + z * 0.05)
phi_wrapped = np.angle(np.exp(1j * phi))
phi_unwrapped = unwrap(phi_wrapped, wrap_around_axis_1=True)
def test_unwrap3D_y_wraparound(self):
"""
Regression test for an incorrect addressing when wrapping around Y-axis is enabled.
Produced a segfault.
"""
x, y, z = np.ogrid[:11, :128, :128]
phi = 2*np.pi*(x*0.2 + y*0.1 + z*0.05)
phi_wrapped = np.angle(np.exp(1j*phi))
phi_unwrapped = unwrap(phi_wrapped, wrap_around_axis_1=True)
def calc_B0_map(folder,roi_mask):
from human_SWI_data import phase_rescale
import unwrap
images,TE=read_dicoms(folder,'EchoTime')
phase_images = [phase_rescale(img) for img in images]
roi_mask_inv=np.zeros_like(roi_mask)
roi_mask_inv[roi_mask==1]=0
roi_mask_inv[roi_mask==0]=1
phase_images_unwrapped=[unwrap.unwrap(np.ma.masked_array(img,roi_mask_inv)) for img in phase_images]
return phase_images_unwrapped
def test_unwrap3D_masked(self):
x, y, z = np.ogrid[:8, :12, :4]
phi = 2*np.pi*(x*0.2 + y*0.1 + z*0.05)
phi_wrapped = np.angle(np.exp(1j*phi))
mask = np.zeros_like(phi, dtype = np.uint8)
mask[4:6, 4:6, 1:3] = 1
phi_wrapped_masked = np.ma.array(phi_wrapped, dtype = np.float32, mask = mask)
phi_unwrapped_masked = unwrap(phi_wrapped_masked)
s = np.round(phi_unwrapped_masked[0,0,0]/(2*np.pi))
assert_array_almost_equal(phi + 2*np.pi*s, phi_unwrapped_masked)
def test_unwrap3D_masked(self):
x, y, z = np.ogrid[:8, :12, :4]
phi = 2 * np.pi * (x * 0.2 + y * 0.1 + z * 0.05)
phi_wrapped = np.angle(np.exp(1j * phi))
mask = np.zeros_like(phi, dtype=np.uint8)
mask[4:6, 4:6, 1:3] = 1
phi_wrapped_masked = np.ma.array(
phi_wrapped, dtype=np.float32, mask=mask)
phi_unwrapped_masked = unwrap(phi_wrapped_masked)
s = np.round(phi_unwrapped_masked[0, 0, 0] / (2 * np.pi))
assert_array_almost_equal(phi + 2 * np.pi * s, phi_unwrapped_masked)
def plot_one_line(med_folder_name, n_column=10):
st_diagram = get_st_diagram(med_folder_name)
fig, ax = plt.subplots(figsize=(12, 8))
gap = st_diagram[:, n_column] - np.unwrap(st_diagram[:, n_column])
st_diagram = (st_diagram.T - gap).T
ax.imshow(unwrap(st_diagram))
# plt.colorbar()
plt.show()
def calc_B0_map(folder, roi_mask):
from human_SWI_data import phase_rescale
import unwrap
images, TE = read_dicoms(folder, 'EchoTime')
phase_images = [phase_rescale(img) for img in images]
roi_mask_inv = np.zeros_like(roi_mask)
roi_mask_inv[roi_mask == 1] = 0
roi_mask_inv[roi_mask == 0] = 1
phase_images_unwrapped = [
unwrap.unwrap(np.ma.masked_array(img, roi_mask_inv))
for img in phase_images
]
return phase_images_unwrapped
def unwrap_plots():
x, y = np.ogrid[:32, :32]
phi = 1 * np.arctan2(x - 14.3, y - 6.3) - 2 * np.arctan2(
x - 18.3, y - 22.1)
#phi[8,8] = np.NaN
phi_wrapped = np.angle(np.exp(1j * phi))
phi_unwrapped = unwrap(
phi_wrapped,
#wrap_around_axis_0 = True,
#wrap_around_axis_1 = True,
)
mask = np.zeros_like(phi, dtype=np.uint8)
mask[10:22, 4:10] = 1
phi_wrapped_masked = np.ma.array(phi_wrapped, dtype=np.float32, mask=mask)
phi_unwrapped_masked = unwrap(phi_wrapped_masked)
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
fig = plt.figure()
plt.gray()
s = fig.add_subplot(221)
s.imshow(phi, interpolation='nearest')
s.set_title('original phase')
s = fig.add_subplot(222)
s.imshow(phi_wrapped, interpolation='nearest')
s.set_title('wrapped phase')
s = fig.add_subplot(223)
s.imshow(phi_unwrapped, interpolation='nearest')
s.set_title('unwrapped phase')
s = fig.add_subplot(224)
s.imshow(phi_unwrapped_masked, interpolation='nearest')
s.set_title('unwrapped phase with mask')
fig.savefig('test.png')
def unwrap_plots():
x, y = np.ogrid[:32, :32]
phi = 1*np.arctan2(x-14.3, y-6.3) - 2*np.arctan2(x-18.3, y-22.1)
#phi[8,8] = np.NaN
phi_wrapped = np.angle(np.exp(1j*phi))
phi_unwrapped = unwrap(phi_wrapped,
#wrap_around_axis_0 = True,
#wrap_around_axis_1 = True,
)
mask = np.zeros_like(phi, dtype = np.uint8)
mask[10:22, 4:10] = 1
phi_wrapped_masked = np.ma.array(phi_wrapped, dtype = np.float32, mask = mask)
phi_unwrapped_masked = unwrap(phi_wrapped_masked)
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
fig = plt.figure()
plt.gray()
s = fig.add_subplot(221)
s.imshow(phi, interpolation = 'nearest')
s.set_title('original phase')
s = fig.add_subplot(222)
s.imshow(phi_wrapped, interpolation = 'nearest')
s.set_title('wrapped phase')
s = fig.add_subplot(223)
s.imshow(phi_unwrapped, interpolation = 'nearest')
s.set_title('unwrapped phase')
s = fig.add_subplot(224)
s.imshow(phi_unwrapped_masked, interpolation = 'nearest')
s.set_title('unwrapped phase with mask')
fig.savefig('test.png')
def sinogram_as_radon(uSin, align=True):
u""" Computes the phase from a complex wave field sinogram.
This step is essential when using the ray approximation before
computation of the refractive index with the inverse Radon
transform.
Parameters
----------
uSin : 2d or 3d complex ndarray
The background-corrected sinogram of the complex scattered wave
:math:`u(\mathbf{r})/u_0(\mathbf{r})`. The first axis iterates
through the angles :math:`\phi_0`.
align : bool
Tries to correct for a phase offset in the phase sinogram.
Returns
-------
phase : 2d or 3d real ndarray
The unwrapped phase array corresponding to ``uSin``.
Notes
-----
The phase-unwrapping is performed with the `unwrap`_ package.
.. _unwrap: https://pypi.python.org/pypi/unwrap
"""
ndims = len(uSin.shape)
if ndims == 2:
# unwrapping is very important
phiR = np.unwrap(np.angle(uSin), axis=-1)
else:
# Unwrap gets the dimension of the problem from the input
# data. Since we have a sinogram, we need to pass it the
# slices one by one.
phiR = np.angle(uSin)
for i in range(len(phiR)):
phiR[i] = unwrap.unwrap(phiR[i])
if align:
align_unwrapped(phiR)
return phiR
def __data_generation(self, list_IDs_temp):
'Generates data containing batch_size samples' # X : (n_samples, *dim, n_channels)
# Initialization
batch_imgs = list()
batch_labels = list()
# Generate data
for i in list_IDs_temp:
# Store sample
# print (self.pair[i][0])
img = scipy.io.loadmat(self.pair[i][0])['wrap']
img_normalized = normalize_angle(img)
batch_imgs.append(img_normalized)
label = unwrap(img,
wrap_around_axis_0=False,
wrap_around_axis_1=False,
wrap_around_axis_2=False)
label_normalized = normalize_angle(label)
batch_labels.append(label_normalized)
return np.array(np.expand_dims(batch_imgs, axis=-1)), np.array(
np.expand_dims(batch_labels, axis=-1))
def calculate_phase_diff_map_1D(dY, dY0, th, ns, mask_for_unwrapping=None):
"""
# % Basic FTP treatment.
# % This function takes a deformed and a reference image and calculates the phase difference map between the two.
# %
# % INPUTS:
# % dY = deformed image
# % dY0 = reference image
# % ns = size of gaussian filter
# %
# % OUTPUT:
# % dphase = phase difference map between images
"""
ny, nx = np.shape(dY)
phase0 = np.zeros([nx,ny])
phase = np.zeros([nx,ny])
for lin in range (0,nx):
fY0=np.fft.fft(dY0[lin,:])
fY=np.fft.fft(dY[lin,:])
dfy=1./ny
fy=np.arange(dfy,1,dfy)
imax=np.argmax(np.abs(fY0[9:int(np.floor(nx/2))]))
ifmax=imax+9
HW=np.round(ifmax*th)
W=2*HW
win=signal.tukey(int(W),ns)
# OJO QUE ACA LO MODIFIQUE
#gaussfilt1D= np.zeros([1,nx])
gaussfilt1D= np.zeros(nx)
gaussfilt1D[int(ifmax-HW-1):int(ifmax-HW+W-1)]=win
# Multiplication by the filter
Nfy0 = fY0*gaussfilt1D
Nfy = fY*gaussfilt1D
# Inverse Fourier transform of both images
Ny0=np.fft.ifft(Nfy0)
Ny=np.fft.ifft(Nfy)
phase0[lin,:] = np.angle(Ny0)
phase[lin,:] = np.angle(Ny)
# 2D-unwrapping is available with masks (as an option), using 'unwrap' library
# unwrap allows for the use of wrapped_arrays, according to the docs:
# "[...] in this case masked entries are ignored during the phase unwrapping process. This is useful if the wrapped phase data has holes or contains invalid entries. [...]"
if mask_for_unwrapping is None:
mphase0 = unwrap(phase0)
mphase = unwrap(phase)
else:
mphase0 = ma.masked_array(phase0, mask=mask_for_unwrapping)
mphase = ma.masked_array(phase, mask=mask_for_unwrapping)
mphase0 = unwrap(mphase0)
mphase = unwrap(mphase)
# Definition of the phase difference map
dphase = (mphase-mphase0);
# dphase = dphase - np.min(dphase) - np.pi/2
return dphase
fB = odt.backpropagate_2d(u_sinB, angles, res, nmed, lD*res)
nB = odt.odt_to_ri(fB, res, nmed)
# Rytov
u_sinR = odt.sinogram_as_rytov(sino/u0)
fR = odt.backpropagate_2d(u_sinR, angles, res, nmed, lD*res)
nR = odt.odt_to_ri(fR, res, nmed)
# Rytov 50
u_sinR50 = odt.sinogram_as_rytov((sino/u0)[::5,:])
fR50 = odt.backpropagate_2d(u_sinR50, angles[::5], res, nmed, lD*res)
nR50 = odt.odt_to_ri(fR50, res, nmed)
# Plot sinogram phase and amplitude
ph = unwrap.unwrap(np.angle(sino/u0))
am = np.abs(sino/u0)
# prepare plot
vmin = np.min(np.array([phantom, nB.real, nR50.real, nR.real]))
vmax = np.max(np.array([phantom, nB.real, nR50.real, nR.real]))
fig, axes = plt.subplots(2,3, figsize=(12,7), dpi=300)
axes = np.array(axes).flatten()
phantommap = axes[0].imshow(phantom, vmin=vmin, vmax=vmax)
axes[0].set_title("phantom \n(non-centered cylinder)")
A = cfg["A"] # number of projections
x = np.arange(size) - size / 2.0
X, Y = np.meshgrid(x, x)
rad_px = radius * res
phantom = np.array(((Y - lC * res)**2 + X**2) < rad_px**2,
dtype=np.float) * (ncyl - nmed) + nmed
u_sinR = odt.sinogram_as_rytov(sino / u0)
# Rytov 200 projections
# remove 50 projections from total of 250 projections
remove200 = np.argsort(angles % .0002)[:50]
angles200 = np.delete(angles, remove200, axis=0)
u_sinR200 = np.delete(u_sinR, remove200, axis=0)
ph200 = unwrap.unwrap(np.angle(sino / u0))
ph200[remove200] = 0
fR200 = odt.backpropagate_2d(u_sinR200, angles200, res, nmed, lD * res)
nR200 = odt.odt_to_ri(fR200, res, nmed)
fR200nw = odt.backpropagate_2d(u_sinR200,
angles200,
res,
nmed,
lD * res,
weight_angles=False)
nR200nw = odt.odt_to_ri(fR200nw, res, nmed)
# Rytov 50 projections
remove50 = np.argsort(angles % .0002)[:200]
angles50 = np.delete(angles, remove50, axis=0)
def PhaseUnwrap(image, noZernikeModes, MIDDLE, DIAMETER, REGION=30):
#convert image to array and float
data = numpy.asarray(image)
datafloat = numpy.zeros((data.shape), dtype=float)
datafloat = data * 1.0
#create the mask to cut out unwanted image areas
mask = numpy.ones(data.shape)
mask = cv2.circle(mask,
center=tuple(MIDDLE),
radius=int(round(DIAMETER / 2)),
color=0,
thickness=-1)
mask = mask - 1
mask = abs(mask)
#mask image to remove extraneous data from edges
datafloat = datafloat * mask
#perform fourier transform and plot
fftarray = numpy.fft.fft2(datafloat, norm='ortho')
fftarray = numpy.fft.fftshift(fftarray)
#remove center section to allow finding of 1st order point
centre = [int(fftarray.shape[1] / 2), int(fftarray.shape[0] / 2)]
order0 = numpy.log((fftarray[centre[1] - REGION:centre[1] + REGION,
centre[0] - REGION:centre[0] + REGION]))
fftarray[centre[1] - REGION:centre[1] + REGION,
centre[0] - REGION:centre[0] + REGION] = 0.00001 + 0j
#find first order point
maxpoint = numpy.argmax(fftarray)
maxpoint = [
int(maxpoint % fftarray.shape[1]),
int(maxpoint / fftarray.shape[1])
]
order1 = ((fftarray[maxpoint[1] - REGION:maxpoint[1] + REGION,
maxpoint[0] - REGION:maxpoint[0] + REGION]))
#pad the fftarray back to original size
order1pad = numpy.zeros((data.shape), dtype=complex)
index = numpy.zeros(4, dtype=int)
index[0] = int((order1pad.shape[1] / 2) - (order1.shape[1] / 2))
index[1] = int((order1pad.shape[1] / 2) + (order1.shape[1] / 2))
index[2] = int((order1pad.shape[0] / 2) - (order1.shape[0] / 2))
index[3] = int((order1pad.shape[0] / 2) + (order1.shape[0] / 2))
order1pad[index[0]:index[1], index[2]:index[3]] = order1
#shift the quadrants back to format for ifft use
order1pad = numpy.fft.ifftshift(order1pad)
#inverse fourier transform
ifftorder1 = numpy.fft.ifft2(order1pad, norm='ortho')
#mask ifftarray to remove extraneous data from edges
ifftorder1 = ifftorder1 * mask
#find phase data by taking 2d arctan of imaginary and real parts
phaseorder1 = numpy.zeros(ifftorder1.shape)
phaseorder1 = numpy.arctan2(ifftorder1.imag, ifftorder1.real)
#mask out edge region to allow unwrap to only use correct region
phasesave = phaseorder1
phaseorder1mask = ma.masked_where(mask == 0, phaseorder1)
#perform unwrap
phaseunwrap = unwrap(phaseorder1mask)
#crop to fill array to improve performance of zernike decomposition
out = numpy.zeros((DIAMETER, DIAMETER), dtype=float)
out = phaseunwrap[MIDDLE[1] - int(round(DIAMETER / 2)):MIDDLE[1] +
int(round(DIAMETER / 2)),
MIDDLE[0] - int(round(DIAMETER / 2)):MIDDLE[0] +
int(round(DIAMETER / 2))]
out = out.filled(0)
print("Calculating Zernike modes")
coef, memLocation1 = opticspy.zernike.fitting(out, noZernikeModes)
return coef, out
def unwrap(self, L_ion): # L_ion is a pointer to an object
unwrap(
L_ion,
[[self.xlo, self.xhi], [self.ylo, self.yhi], [self.zlo, self.zhi]])
def calculate_phase_map(xt, ns=15, roll=True):
"""
Obtains phase map from a spatiotemporal diagram.
Parameters
----------
xt : array
Spatiotemporal diagram
ns : int, optional
Size of the gaussian filter (default is 15)
roll: bool, optional
If True, centers the signal before perfoeming the inverse Fourier Transform to obtain the phase
Returns
-------
phase : array
Phase map
"""
phase = np.zeros_like(xt, dtype='float')
ft = np.fft.fft(xt, axis=1)
ny, nx = np.shape(ft)
th = 0.9
num = 9
max_array = np.zeros(xt.shape[0])
for i in range(xt.shape[0]):
ftlin = ft[i,:]
imax=np.argmax(np.abs(ftlin[num:int(np.floor(nx/2))]))+num
max_array[i] = imax
ifmax = int(np.mean(max_array))
HW=np.round(ifmax*th)
W=2*HW
win=signal.tukey(int(W),ns)
gaussfilt1D= np.zeros(nx)
gaussfilt1D[int(ifmax-HW-1):int(ifmax-HW+W-1)]=win
for i in range(xt.shape[0]):
ftlin = ft[i,:]
# Multiplication by the filter
Nfy = ftlin*gaussfilt1D
if roll==True:
c_centered = np.roll(Nfy, -ifmax)
else:
c_centered = Nfy
# Inverse Fourier transform
Cxt = np.fft.ifft(c_centered)
# 1D unwrapping
phi = np.unwrap(np.imag(np.log(Cxt)))
phase[i,:] = phi
# 2D unwrapping
phase = unwrap(phase)
# Mean substraction
for i in range(xt.shape[0]):
line_phase = np.unwrap(phase[i,:])
phase[i,:] = line_phase-np.mean(line_phase)
return phase
def download_c(directory):
directory += '-c'
download(directory, relative_path('c.txt'))
clean(directory, ['c', 'h'])
unwrap(directory)
for i in range(start, end):
pairs.append([('wrap/wraped_' + str(i + 1) + '.mat'),
('unwrap/un_wraped_' + str(i + 1) + '.mat')])
return pairs
from random import shuffle
test_pair = make_pair(3750, 4000)
from random import sample, choice
temp = choice(test_pair)
print(temp[0], "GAB", temp[1])
import matplotlib.pyplot as plt
img = scipy.io.loadmat(temp[0])['wrap']
mask_x = unwrap(img,
wrap_around_axis_0=False,
wrap_around_axis_1=False,
wrap_around_axis_2=False)
plt.figure(figsize=(10, 10))
plt.subplot(121)
plt.imshow(dg.normalize_angle(img), cmap='jet')
plt.subplot(122)
plt.imshow(dg.normalize_angle(mask_x), cmap='jet')
plt.show()
class_map = 1
test_generator = dg.DataGenerator(test_pair,
class_map,
batch_size=20,
dim=(256, 256, 1),
shuffle=True)
test_steps = test_generator.__len__()
test_steps
from example_helper import load_cell
# load initial cell
cell1 = load_cell("HL60_field.zip")
# refocus to two different positions
cell2 = nrefocus.refocus(cell1, 15, 1, 1) # forward
cell3 = nrefocus.refocus(cell1, -15, 1, 1) # backward
# amplitude range
vmina = np.min(np.abs(cell1))
vmaxa = np.max(np.abs(cell1))
ampkw = {"cmap": plt.get_cmap("gray"), "vmin": vmina, "vmax": vmaxa}
# phase range
cell1p = unwrap.unwrap(np.angle(cell1))
cell2p = unwrap.unwrap(np.angle(cell2))
cell3p = unwrap.unwrap(np.angle(cell3))
vminp = np.min(cell1p)
vmaxp = np.max(cell1p)
phakw = {"cmap": plt.get_cmap("coolwarm"), "vmin": vminp, "vmax": vmaxp}
# plots
fig, axes = plt.subplots(2, 3, figsize=(8, 4.5))
axes = axes.flatten()
for ax in axes:
ax.xaxis.set_major_locator(plt.NullLocator())
ax.yaxis.set_major_locator(plt.NullLocator())
# titles
axes[0].set_title("focused backward")
def unwrap_3D(img, roi_mask):
import unwrap
img_unwrapped = unwrap.unwrap(np.ma.masked_array(img, roi_mask))
return img_unwrapped
def download_python(directory):
directory += '-python'
download(directory, relative_path('python.txt'))
clean(directory, ['py'])
unwrap(directory)
def download_java(directory):
directory += '-java'
download(directory, relative_path('java.txt'))
clean(directory, ['java'])
unwrap(directory)
def sinogram_as_rytov(uSin, u0=1, align=True):
u""" Converts the complex wave field sinogram to Rytov data
This method applies the Rytov approximation to the
recorded complex wave sinogram. To achieve this, the following
filter is applied:
.. math::
u_\mathrm{B}(\mathbf{r}) = u_\mathrm{0}(\mathbf{r})
\ln\!\\left(
\\frac{u_\mathrm{R}(\mathbf{r})}{u_\mathrm{0}(\mathbf{r})}
+1 \\right)
This filter step effectively replaces the Born approximation
:math:`u_\mathrm{B}(\mathbf{r})` with the Rytov approximation
:math:`u_\mathrm{R}(\mathbf{r})`, assuming that the scattered
field is equal to
:math:`u(\mathbf{r})\\approx u_\mathrm{R}(\mathbf{r})+
u_\mathrm{0}(\mathbf{r})`.
Parameters
----------
uSin : 2d or 3d complex ndarray
The sinogram of the complex wave
:math:`u_\mathrm{R}(\mathbf{r}) + u_\mathrm{0}(\mathbf{r})`.
The first axis iterates through the angles :math:`\phi_0`.
u0 : ndarray of dimension as ``uSin`` or less, or int.
The incident plane wave
:math:`u_\mathrm{0}(\mathbf{r})` at the detector.
If ``u0`` is "1", it is assumed that the data is already
background-corrected (
``uSin`` :math:`= \\frac{u_\mathrm{R}(\mathbf{r})}{
u_\mathrm{0}(\mathbf{r})} + 1`
). Note that if the reconstruction distance :math:`l_\mathrm{D}`
of the original experiment is non-zero and `u0` is set to 1,
then the reconstruction will be wrong; the field is not focused
to the center of the reconstruction volume.
align : bool
Tries to correct for a phase offset in the phase sinogram.
Returns
-------
uB : 2d or 3d real ndarray
The Rytov-filtered complex sinogram
:math:`u_\mathrm{B}(\mathbf{r})`.
Notes
-----
The phase-unwrapping is performed with the `unwrap`_ package.
.. _unwrap: https://pypi.python.org/pypi/unwrap """
ndims = len(uSin.shape)
# imaginary part of the complex Rytov phase
phiR = np.angle(uSin / u0)
# real part of the complex Rytov phase
lna = np.log(np.absolute(uSin / u0))
if ndims == 2:
# unwrapping is very important
phiR[:] = np.unwrap(phiR, axis=-1)
else:
# Unwrap gets the dimension of the problem from the input
# data. Since we have a sinogram, we need to pass it the
# slices one by one.
for i in range(len(phiR)):
phiR[i] = unwrap.unwrap(phiR[i])
if align:
align_unwrapped(phiR)
#rytovSin = u0*(np.log(a/a0) + 1j*phiR)
# u0 is one - we already did background correction
# complex rytov phase:
rytovSin = 1j * phiR + lna
return u0 * rytovSin
A = cfg["A"] # number of projections
#phantom = np.loadtxt(arc.open("mie_phantom.txt"))
x = np.arange(size)-size/2.0
X,Y = np.meshgrid(x,x)
rad_px = radius*res
phantom = np.array(((Y-lC*res)**2+X**2)<rad_px**2, dtype=np.float)*(ncyl-nmed)+nmed
u_sinR = odt.sinogram_as_rytov(sino/u0)
# Rytov 200 projections
# remove 50 projections from total of 250 projections
remove200 = np.argsort(angles % .002)[:50]
angles200 = np.delete(angles, remove200, axis=0)
u_sinR200 = np.delete(u_sinR, remove200, axis=0)
ph200 = unwrap.unwrap(np.angle(sino/u0))
ph200[remove200] = 0
fR200 = odt.backpropagate_2d(u_sinR200, angles200, res, nmed, lD*res)
nR200 = odt.odt_to_ri(fR200, res, nmed)
fR200nw = odt.backpropagate_2d(u_sinR200, angles200, res, nmed, lD*res, weight_angles=False)
nR200nw = odt.odt_to_ri(fR200nw, res, nmed)
# Rytov 50 projections
remove50 = np.argsort(angles % .002)[:200]
angles50 = np.delete(angles, remove50, axis=0)
u_sinR50 = np.delete(u_sinR, remove50, axis=0)
ph50 = unwrap.unwrap(np.angle(sino/u0))
ph50[remove50] = 0
+'Procesado FTP: '+str(round(tiempo_ftp/60,2))+' minutos'+'\n' \
+'Coordenadas polares: '+str(round(tiempo_coord_polares/60,2))+' minutos'+'\n' \
+'Tiempo total: '+str(round(round(tiempo_hdf5/60,2)+round(tiempo_ftp/60,2)+round(tiempo_coord_polares/60,2))) +' minutos'
# mensaje = str(filename) +'\n' \
# +'Procesado FTP: '+str(round(tiempo_ftp/60,2))+' minutos'+'\n' \
# +'Coordenadas polares: '+str(round(tiempo_coord_polares/60,2))+' minutos'+'\n' \
# +'Tiempo total: '+str(round(round(tiempo_ftp/60,2)+round(tiempo_coord_polares/60,2))) +' minutos'
print(mensaje)
f = h5py.File(destination_path+filename+'-annulus-PRO.hdf5','r')
p = f['height_fields/annulus_polar']
fps = 250
R1, R2 = (205/2)/10, (215/2)/10 #en cm
Rprom = (R1+R2)/2
xfinal = 2*np.pi*Rprom
plt.figure()
plt.imshow(unwrap(p), extent=[0,xfinal, 0, p.shape[1]/fps], aspect='auto',cmap='RdYlBu')
plt.clim((-1,1))
plt.colorbar()
plt.xlabel('longitud de arco (cm)')
plt.ylabel('tiempo (s)')
plt.title(filename)
plt.tight_layout()
plt.savefig(path_mediciones+'graficos/'+filename+'.png')
bot.send_photo(chat_id=my_id, photo=open(path_mediciones+'graficos/'+filename+'.png', 'rb'), caption=mensaje)
#bot.send_photo(chat_id=pablo_id, photo=open(path_mediciones+'graficos/'+filename+'.png', 'rb'), caption=mensaje)
def unwrap2D(wraped_phase, type="boundary", noise=True):
"""
2D unwarp function. There are several type to
use in several different situation.
Type:
-----------------------------------------------
Simple: The very simple algorithm to unwrap 2D warpped phase
just scan the whole matrix to unwrap phase. Very noise
sensitive
boundary: 2D phase unwrap method for phase map with aperture,
for example, circle, rectangular, ring or slit. It
use a DFS(deep first search) algorithm to traverse
the phase map, then unwarp the phase.It is also very
noise sensitive
etc: still have more, to be continue
"""
if type == "simple" and noise == False:
l = len(wraped_phase)
b = []
b = np.array(b)
for i in range(l):
if i % 2 == 0:
b = np.append(b, wraped_phase[i])
else:
b = np.append(b, wraped_phase[i][::-1])
ph1 = unwrap1D(b)
ph = np.zeros([l, l])
for i in range(l):
if i % 2 == 0:
ph[i] = ph1[0:l]
ph1 = ph1[l:]
else:
ph[i] = ph1[0:l][::-1]
ph1 = ph1[l:]
return ph
elif type == "boundary" and noise == False:
ph1 = wraped_phase[0]
M = wraped_phase[1]
s = wraped_phase[2]
start_pixel = np.where(M == 1)
m = start_pixel[0][0]
n = start_pixel[1][0]
print("start pixel", m, n)
ph = DFS(M, ph1, m, n, s)
return ph
elif noise == True:
ph1 = wraped_phase[0]
M = wraped_phase[1]
s = wraped_phase[2]
ph = unwrap(ph1,wrap_around_axis_0=False,\
wrap_around_axis_1=False,\
wrap_around_axis_2=False)
return ph
else:
print("No this type of unwrap algorithm")
return 0
def unwrap_3D(img, roi_mask):
import unwrap
img_unwrapped=unwrap.unwrap(np.ma.masked_array(img,roi_mask))
return img_unwrapped
res = cfg["res"] # px/wavelengths
A = cfg["A"] # number of projections
#phantom = np.loadtxt(arc.open("mie_phantom.txt"))
x = np.arange(size)-size/2.0
X,Y = np.meshgrid(x,x)
rad_px = radius*res
phantom = np.array(((Y-lC*res)**2+X**2)<rad_px**2, dtype=np.float)*(ncyl-nmed)+nmed
u_sinR = odt.sinogram_as_rytov(sino/u0)
# Rytov 100 projections evenly distributed
removeeven = np.argsort(angles % .002)[:150]
angleseven = np.delete(angles, removeeven, axis=0)
u_sinReven = np.delete(u_sinR, removeeven, axis=0)
pheven = unwrap.unwrap(np.angle(sino/u0))
pheven[removeeven] = 0
fReven = odt.backpropagate_2d(u_sinReven, angleseven, res, nmed, lD*res)
nReven = odt.odt_to_ri(fReven, res, nmed)
fRevennw = odt.backpropagate_2d(u_sinReven, angleseven, res, nmed, lD*res, weight_angles=False)
nRevennw = odt.odt_to_ri(fRevennw, res, nmed)
# Rytov 100 projections more than 180
removemiss = 249 - np.concatenate((np.arange(100), 100+np.arange(150)[::3]))
anglesmiss = np.delete(angles, removemiss, axis=0)
u_sinRmiss = np.delete(u_sinR, removemiss, axis=0)
phmiss = unwrap.unwrap(np.angle(sino/u0))
phmiss[removemiss] = 0
