def return_grad(data, filter_grad=False, grad_filter_max=0.5, gfsize=0):
#r_dx = _np.zeros_like(r)
#r_dx[1:,:] = r[1:,:]-r[:-1,:]
#r_dy = _np.zeros_like(r)
#r_dy[:,1:] = r[:,1:]-r[:,:-1]
data_dx = _np.zeros_like(data)
data_dx[1:, :] = -1 * (data[1:, :] - data[:-1, :])
data_dy = _np.zeros_like(data)
data_dy[:, 1:] = (data[:, 1:] - data[:, :-1])
data_dxdy = -1 * _np.array([data_dx, data_dy])
if filter_grad:
wx = _np.where(abs(data_dxdy[0]) > grad_filter_max)
wy = _np.where(abs(data_dxdy[1]) > grad_filter_max)
#print "Length wx: ", len(wx[0])
'''
for i in range(len(wx[0])):
wx0 = wx[0][i]
wx1 = wx[1][i]
surrounding = data_dxdy[0][wx0-1:wx0+2,wx1-1:wx1+2].mean()
data_dxdy[0][wx0,wx1] = surrounding
for i in range(len(wy[0])):
wy0 = wy[0][i]
wy1 = wy[1][i]
surrounding = data_dxdy[1][wy0-1:wy0+2,wy1-1:wy1+2].mean()
data_dxdy[1][wy0,wy1] = surrounding
'''
data_dxdy[0][wx[0], wx[1]] = 0
data_dxdy[1][wy[0], wy[1]] = 0
if gfsize > 0:
data_dxdy[0] = gf(data_dxdy[0], gfsize)
data_dxdy[1] = gf(data_dxdy[1], gfsize)
return data_dxdy
def return_grad(data, filter_grad=False, grad_filter_max=0.5, gfsize=0):
#r_dx = _np.zeros_like(r)
#r_dx[1:,:] = r[1:,:]-r[:-1,:]
#r_dy = _np.zeros_like(r)
#r_dy[:,1:] = r[:,1:]-r[:,:-1]
data_dx = _np.zeros_like(data)
data_dx[1:,:] = -1*(data[1:,:]-data[:-1,:])
data_dy = _np.zeros_like(data)
data_dy[:,1:] = (data[:,1:]-data[:,:-1])
data_dxdy = -1*_np.array([data_dx,data_dy])
if filter_grad:
wx = _np.where(abs(data_dxdy[0])>grad_filter_max)
wy = _np.where(abs(data_dxdy[1])>grad_filter_max)
#print "Length wx: ", len(wx[0])
'''
for i in range(len(wx[0])):
wx0 = wx[0][i]
wx1 = wx[1][i]
surrounding = data_dxdy[0][wx0-1:wx0+2,wx1-1:wx1+2].mean()
data_dxdy[0][wx0,wx1] = surrounding
for i in range(len(wy[0])):
wy0 = wy[0][i]
wy1 = wy[1][i]
surrounding = data_dxdy[1][wy0-1:wy0+2,wy1-1:wy1+2].mean()
data_dxdy[1][wy0,wy1] = surrounding
'''
data_dxdy[0][wx[0],wx[1]] = 0
data_dxdy[1][wy[0],wy[1]] = 0
if gfsize>0:
data_dxdy[0] = gf(data_dxdy[0],gfsize)
data_dxdy[1] = gf(data_dxdy[1],gfsize)
return data_dxdy
def findcenter(self, smooth_sigma=1.8, smoothbg_sigma=40):
"""
Finds the pixel with the highest intensity.
Parameters:
-----------
smooth_sigma : number
Sigma for the 2D gaussian that is to convolve
the frame before attempting to find the center.
smoothbg_sigma : number
Sigma for the 2D gaussian that will be used
to remove low frequency variations across the
detector.
"""
try:
arr = self.diff
except AttributeError:
arr = self.diff_savesets()
nframes = len(arr)
locs = reshape(zeros(nframes * 2, dtype='int'), (nframes, 2))
for i in range(nframes):
a = arr[i]
b = gf(a, smooth_sigma) - gf(a, smoothbg_sigma)
locs[i] = [j[0] for j in where(b == b.max())]
self.centerlocs = locs
return locs
def color_diffuse(self, amount):
sigma = 1.5 * amount + 2
scale = amount
labout = np.copy(self.labin)
labout[:, :, 1] = gf(self.labin[:, :, 1], sigma)
labout[:, :, 2] = gf(self.labin[:, :, 2], sigma)
diffuse = cv2.cvtColor(labout, cv2.COLOR_LAB2RGB)
return diffuse
def random_gal(galreal,Nrand,Nmin=10):
""" Prefered random galaxy generator. For a given galaxy with a
given magnitude (and other properties), it calculates the redshift
sensitivity function from galaxies in a magnitude band around the
selected one (i.e., including slightly brighter and fainter
galaxies), and places Nrand new galaxies at a random redshift given
by a smoothed version of the observed sensitivity function. For
extremely bright or faint galaxies (rare) the sensitivity function
is calculated from at least Nmin (=50) galaxies (i.e. the magnitude
band is increased)."""
from astro.sampledist import RanDist
from astro.fit import InterpCubicSpline
from scipy.ndimage import gaussian_filter as gf
Ckms = 299792.458
Nmin = int(Nmin) #minimum number of galaxys for the fit
zmin = np.min(galreal.ZGAL)
zmax = np.max(galreal.ZGAL) + 0.1
DZ = 0.01 #delta z for the histogram for getting the spline in z
smooth_scale = 10. #smoothing scale for the histogram (in number
#of bins, so depends on DZ)
galreal.sort(order='MAG') #np.recarray.sort()
galrand = galreal.repeat(Nrand)
delta_mag = 0.5 # half of the magnitude bandwidth to generate the z histogram
bins = np.append(np.linspace(0,zmin,20),np.arange(zmin+DZ, zmax, DZ))
for i in xrange(len(galreal)):
if i < Nmin:
vals,bins = np.histogram(galreal.ZGAL[:Nmin], bins)
vals = gf(vals.astype(float),smooth_scale) # smooth the histogram
spl = InterpCubicSpline(0.5*(bins[:-1] + bins[1:]), vals.astype(float))
else:
delta_mag2=delta_mag
while True:
cond = (galreal.MAG > 0) & (galreal.MAG < 90)
cond = cond & (galreal.MAG<=galreal.MAG[i]+delta_mag2)&(galreal.MAG>galreal.MAG[i]-delta_mag2)
if np.sum(cond)>=Nmin:
break
else:
delta_mag2+=0.1
vals,bins = np.histogram(galreal.ZGAL[cond], bins)
vals = gf(vals.astype(float),smooth_scale) # smooth the histogram
spl = InterpCubicSpline(0.5*(bins[:-1] + bins[1:]), vals.astype(float))
rvals = np.linspace(0, zmax, 1e4)
rand_z = RanDist(rvals, spl(rvals))
zrand = rand_z.random(Nrand)
integer_random2 = np.random.randint(0,len(galreal),Nrand) #for RA,DEC
RArand = galreal.RA[integer_random2]
DECrand = galreal.DEC[integer_random2]
galrand.ZGAL[i*Nrand:(i+1)*Nrand] = zrand
galrand.RA[i*Nrand:(i+1)*Nrand] = RArand
galrand.DEC[i*Nrand:(i+1)*Nrand] = DECrand
return galrand
def random_gal(galreal,Nrand,Nmin=20):
""" Prefered random galaxy generator. For a given galaxy with a
given magnitude (and other properties), it calculates the redshift
sensitivity function from galaxies in a magnitude band around the
selected one (i.e., including slightly brighter and fainter
galaxies), and places Nrand new galaxies at a random redshift given
by a smoothed version of the observed sensitivity function. For
extremely bright or faint galaxies (rare) the sensitivity function
is calculated from at least Nmin (=50) galaxies (i.e. the magnitude
band is increased)."""
from astro.sampledist import RanDist
from astro.fit import InterpCubicSpline
from scipy.ndimage import gaussian_filter as gf
Ckms = 299792.458
Nmin = int(Nmin) #minimum number of galaxys for the fit
zmin = np.min(galreal.ZGAL)
zmax = np.max(galreal.ZGAL) + 0.1
DZ = 0.01 #delta z for the histogram for getting the spline in z
smooth_scale = 10. #smoothing scale for the histogram (in number
#of bins, so depends on DZ)
galreal.sort(order='MAG') #np.recarray.sort()
galrand = galreal.repeat(Nrand)
delta_mag = 0.5 # half of the magnitude bandwidth to generate the z histogram
bins = np.append(np.linspace(0,zmin,20),np.arange(zmin+DZ, zmax, DZ))
for i in xrange(len(galreal)):
if i < Nmin:
vals,bins = np.histogram(galreal.ZGAL[:Nmin], bins)
vals = gf(vals.astype(float),smooth_scale) # smooth the histogram
spl = InterpCubicSpline(0.5*(bins[:-1] + bins[1:]), vals.astype(float))
else:
delta_mag2=delta_mag
while True:
cond = (galreal.MAG > 0) & (galreal.MAG < 90)
cond = cond & (galreal.MAG<=galreal.MAG[i]+delta_mag2)&(galreal.MAG>galreal.MAG[i]-delta_mag2)
if np.sum(cond)>=Nmin:
break
else:
delta_mag2+=0.1
vals,bins = np.histogram(galreal.ZGAL[cond], bins)
vals = gf(vals.astype(float),smooth_scale) # smooth the histogram
spl = InterpCubicSpline(0.5*(bins[:-1] + bins[1:]), vals.astype(float))
rvals = np.linspace(0, zmax, 1e4)
rand_z = RanDist(rvals, spl(rvals))
zrand = rand_z.random(Nrand)
galrand.ZGAL[i*Nrand:(i+1)*Nrand] = zrand
return galrand
def highlight_feature(cal_stack):
diff = []
grad = []
for i in range(len(cal_stack) - 1):
dd = cal_stack[i] - cal_stack[i + 1]
diff.append(dd)
dd = get_grad(cal_stack[i])
grad.append(dd)
diff = np.array(diff).mean(axis=0)
grad = np.array(grad).mean(axis=0)
featuremap = gf(diff, 5) * gf(grad, 5)
return featuremap
def xi_ag_LS(self,sigma=0,jacknife=True,f1=None,f2=None,f3=None):
""" Returns the abs-gal 2D2PCF using Landy & Szalay
estimator. It smooths the pair-pair counts with a Gaussian
kernel with sigma = [rs,ts] (for each direction, see
scipy.ndimage.gaussian_filter), before computing the
cross-correlation. It also gives the projected along the LOS
measurement from DxDy_1D values (not implemented yet)."""
from pyntejos.xcorr.xcorr import W3
from scipy.ndimage import gaussian_filter as gf
s = sigma
Wag,_ = W3(gf(self.DaDg,s),gf(self.RaRg,s),gf(self.DaRg,s),gf(self.RaDg,s),f1=f1,f2=f2,f3=f3)
#jacknife error
err_Wjk = np.zeros((len(self.rbinedges) - 1, len(self.tbinedges) - 1), float)
if jacknife:
for field in self.fields:
DaDg_aux = self.DaDg - field.DaDg(self.rbinedges,self.tbinedges)
RaRg_aux = self.RaRg - field.RaRg(self.rbinedges,self.tbinedges)
DaRg_aux = self.DaRg - field.DaRg(self.rbinedges,self.tbinedges)
RaDg_aux = self.RaDg - field.RaDg(self.rbinedges,self.tbinedges)
Wag_aux,_ = W3(gf(DaDg_aux,s),gf(RaRg_aux,s),gf(DaRg_aux,s),gf(RaDg_aux,s),f1=f1,f2=f2,f3=f3)
err_Wjk += (Wag - Wag_aux)**2
N = len(self.fields)
err_Wjk = (N - 1.) / N * err_Wjk
err_Wjk = np.sqrt(err_Wjk)
return Wag, err_Wjk
def resistive(run, time, comp, smooth='no',sigma=1):
"""
returns the resistive term of the Ohm's law,
at time if time is float or averaged on the list of time if time is a list (of floats)
@return : np.ndarray of shape run.vfield_shape
"""
if isinstance(time, collections.Iterable) == False:
time = [time]
ntot = 1
else:
ntot = time.size
term = np.zeros(shape = run.sfield_shape, dtype = run.dtype)
for t in time:
term += run.getJ(t, comp)
term *= run.getResistivity()/ntot
if smooth.lower() == 'yes':
term = gf(term, sigma = sigma, order = 0)
return term
def pdderiv(ar, dx=1., ax=0, order=4, smth=None):
"""
pderiv gives the double partial derivative
of a periodic array along a given axis.
Inputs:
ar - The input array
dx - Grid spacing, defaults to 1.
ax - Axis along which to take the derivative
order - Order of accuracy, (2,4) defaults to 2
Output:
dar - The derivative array
"""
if smth is not None:
ar = gf(ar, sigma=smth)
if order == 2:
dar = (np.roll(ar, -1, axis=ax) - 2 * ar +
np.roll(ar, 1, axis=ax)) / dx**2
elif order == 4:
dar = (-np.roll(ar, -2, axis=ax) + 16 * np.roll(ar, -1, axis=ax) -
30 * ar + 16 * np.roll(ar, 1, axis=ax) -
np.roll(ar, 2, axis=ax)) / (12 * dx**2)
return dar
def Velpad(vel, nPMLs, factor):
from scipy.ndimage import gaussian_filter as gf
velPML = np.pad(vel, [[nPMLs[0], nPMLs[1]], [nPMLs[2], nPMLs[3]]], 'edge')
velPML = np.pad(vel, [nPMLs[:2], nPMLs[2:]], 'edge')
Nz, Nx = np.shape(velPML)
for iz in range(nPMLs[0]):
velPML[iz, :] = gf(velPML[iz, :], (nPMLs[0] - iz) * factor)
for iz in range(nPMLs[1]):
velPML[Nz - iz - 1, :] = gf(velPML[Nz - iz - 1, :],
(nPMLs[1] - iz) * factor)
for ix in range(nPMLs[2]):
velPML[:, ix] = gf(velPML[:, ix], (nPMLs[2] - ix) * factor)
for ix in range(nPMLs[3]):
velPML[:, Nx - ix - 1] = gf(velPML[:, Nx - ix - 1],
(nPMLs[3] - ix) * factor)
return velPML
def _load_movie(self, time, slc):
mvars = 'all'
d = self._M.get_fields('all', time, slc=slc)
# This is dumb and should be done in moive
#if slc is None:
# ip,jp = 2*[np.s_[:]]
#else:
# ip,jp = slc[:2]
#d['xx'] = d['xx'][ip]
#d['yy'] = d['yy'][jp]
for q,s in zip([1.,-1.],'ie'):
for k in 'xx xy xz yy yz zz'.split():
d['t'+s+k] = d['p'+s+k]/d['n'+s]
rotate_ten(d,'t'+s, av='')
for k in 'xyz':
d['v'+s+k] = q*d['j'+s+k]/d['n'+s]
mag = lambda _x,_y: _x**2 + _y**2
#d['|b|'] = np.sqrt(reduce(mag, (d['b'+k] for k in'xyz')))
d['|b|'] = np.sqrt(d['bx']**2 + d['by']**2 + d['bz']**2)
d['psi'] = gf(calc_psi(d), sigma=self.sig)
return d
def __fig2pixel(p, plotPxl=False, smoothing=0., LIMITS=XYLIM):
# smoothing is std of gaussian 2d filter. set to 0 to not smooth.
# https://stackoverflow.com/questions/43363388/how-to-save-a-greyscale-matplotlib-plot-to-numpy-array
from skimage import color
ax = plot(p, LIMITS=LIMITS)
fig = ax.get_figure()
fig.canvas.draw()
ax.axis("off")
width, height = fig.get_size_inches() * fig.get_dpi()
# print("dpi: {}".format(fig.get_dpi()))
# import pdb
# pdb.set_trace()
img = np.frombuffer(fig.canvas.tostring_rgb(),
dtype='uint8').reshape(int(height), int(width), 3)
img = color.rgb2gray(img)
if smoothing > 0:
img = gf(img, smoothing, truncate=5)
if plotPxl:
# - show the figure
plt.figure()
# plt.imshow(img, vmin=0, vmax=1, cmap="gray", interpolation="bicubic")
plt.imshow(img, vmin=0, vmax=1, cmap="gray")
return img
def collapse_along_LOS(DD, nbins=None, s=0):
"""Sums pair counts over the first nbins along the sightline
dimension. Returns an array with the values for transverse bins. If
nbins is None then collapses the whole array.
Parameters
----------
DD : ndarray
Pair counts to sum
nbins : int, optional
Number of bins in the radial dimension to collapse along
If None, take them all
s : float, optional
For Gaussian filtering
Returns
-------
DD_1D : ndarray
Collapsed pair counts
"""
# Gaussian filter
if s > 0:
sDD = gf(DD, s)
else:
sDD = DD
if nbins is None:
nbins = sDD.shape[0]
# Avoidable loop?
#old_DD_1D = np.array([np.sum(sDD.T[i][:nbins]) for i in range(sDD.shape[1])])
DD_1D = np.sum(sDD[:nbins, :], axis=0)
# Return
return DD_1D
def __init__(self, cell, ch):
self.name = cell
movie = Image.open(self.name + '.tif')
self.ch = ch[0]
self.sample = ch[1]
self.n_ch = len(ch[0])
self.n_frame = int((movie.n_frames) / self.n_ch)
self.width = movie.width
self.height = movie.height
# I[channel,frame,row,column]
self.I = np.zeros((self.n_ch, self.n_frame, self.height, self.width),
dtype=float)
self.Is = np.zeros((self.n_ch, self.n_frame, self.height, self.width),
dtype=float)
self.mask1 = np.zeros(
(self.n_ch, self.n_frame, self.height, self.width), dtype=bool)
self.mask2 = np.zeros(
(self.n_ch, self.n_frame, self.height, self.width), dtype=bool)
self.Im1 = np.zeros((self.n_ch, self.n_frame, self.height, self.width),
dtype=float)
self.Im2 = np.zeros((self.n_ch, self.n_frame, self.height, self.width),
dtype=float)
self.pcc1 = np.zeros((self.n_ch, self.n_frame), dtype=float)
self.pcc2 = np.zeros((self.n_ch, self.n_frame), dtype=float)
# Read data from each channel
for i in range(self.n_ch):
movie_i = Image.open(cell + '-' + ch[0][i] + '.tif')
# Read data from each frame
for j in range(self.n_frame):
movie_i.seek(j) # Move to frame j
I0 = np.array(movie_i, dtype=float)
self.I[i, j] = I0 - I0.min()
self.Is[i, j] = gf(self.I[i, j], sigma)
# Normalize signal from each channel
for i in range(self.n_ch):
Imax = self.I[i].max()
self.I[i] = self.I[i] / Imax
self.Is[i] = self.Is[i] / Imax
# Masking to find cell boundary and clusters
for j in range(self.n_frame): # Each frame
for i in range(self.n_ch): # Each channel
# Mask1 to find a cell from ch2 (because LAT/LCK are more uniformly spread)
m1 = self.Is[1, j] > np.percentile(self.Is[1, j], p1)
self.mask1[i, j] = m1
self.Im1[i, j, m1] = self.I[i, j, m1]
# Mask2 to find clusters within the cell boundary
self.mask2[i, j, m1] = self.Is[i, j, m1] > np.percentile(
self.Is[i, j, m1], p2)
self.Im2[i, j, self.mask2[i, j]] = self.I[i, j, self.mask2[i,
j]]
i1 = i
i2 = (i1 + 1) % self.n_ch
self.pcc1[i, j] = st.pearsonr(self.I[i1, j, m1].flatten(),
self.I[i2, j, m1].flatten())[0]
def _plot1D(self,
ip,
var_labels=None,
ptargs=None,
ldgargs=None,
xlim=None,
ylim=None):
""" Function to make a plot of a 1D cut
Parameters
=========
ax : (matplotlib.pyplot.axis)
xy : (numpy.array)
vrs : [numpy.array]
Todo: Fix the stuipd kwargs
"""
if var_labels is None: var_labels = self.page_vars_1D
# We will come back it this if we have to
#if not ptargs:
# ptargs = 3*({},)
#elif type(ptargs) is dict:
# ptargs =3*(ptargs,)
_xy = self.d[2 * self._cut_dir]
_lim = _xy[[0, -1]]
_cut_index = np.s_[ip, :]
if self._cut_dir is 'x': _cut_index = _cut_index[::-1]
lines = []
for a, vrs, labs in zip(self.ax, self.page_vars_1D, var_labels):
_sl = []
#for v,l,pwargs in zip(vrs, var_labels, plot_kwargs):
for v, l in zip(vrs, labs):
gfv = gf(self.d[v][_cut_index], sigma=self.sig)
_sl += a.plot(_xy, gfv, label=l)
lines += [_sl]
a.set_xlim(_lim)
_tl = 'cut @ {} = {:1.2f}'.format(
self._not_cut_dir, self.d[2 * self._not_cut_dir][ip])
a.set_title(_tl, size=6, loc='right')
if ldgargs:
if type(ldgargs) is not dict: ldgargs = {}
a.legend(**ldgargs)
a.minorticks_on()
return lines
def VelPadding(self):
from scipy.ndimage import gaussian_filter as gf
nPMLs = self.nPMLs
factor = self.factor
vel = self.vel
velpad = np.pad(vel, [nPMLs[:2], nPMLs[2:]], 'edge')
Nz, Nx = np.shape(velpad)
for iz in range(nPMLs[0]):
velpad[iz, :] = gf(velpad[iz, :], (nPMLs[0] - iz) * factor)
for iz in range(nPMLs[1]):
velpad[Nz - iz - 1, :] = gf(velpad[Nz - iz - 1, :],
(nPMLs[1] - iz) * factor)
for ix in range(nPMLs[2]):
velpad[:, ix] = gf(velpad[:, ix], (nPMLs[2] - ix) * factor)
for ix in range(nPMLs[3]):
velpad[:, Nx - ix - 1] = gf(velpad[:, Nx - ix - 1],
(nPMLs[3] - ix) * factor)
velpad = np.floor(velpad)
return velpad
def phase_corr(fixed, moving, sigma):
if fixed.shape > moving.shape:
print('fixed image is larger than moving', fixed.shape, moving.shape)
fixed = fixed[tuple(map(slice, moving.shape))]
print('fixed image resized to', fixed.shape)
elif fixed.shape < moving.shape:
print('fixed image is smaller than moving', fixed.shape, moving.shape)
moving = moving[tuple(map(slice, fixed.shape))]
print('moving image resized to', moving.shape)
fixed = gf(fixed, sigma=sigma)
moving = gf(moving, sigma=sigma)
print('applying phase correlation')
try:
for i in [0]:
shift, error, diffphase = corr(fixed, moving)
except:
for i in [0]:
shift, error, diffphase = np.zeros(len(moving)), 0, 0
print("couldn't perform PhaseCorr, so shift was casted as zeros")
return shift
def get_dens_from_hab(f: pd.DataFrame):
lam = f['LAT'].min()
laM = f['LAT'].max()
lom = f['LON'].min()
loM = f['LON'].max()
R = .05
lam = np.floor(lam / R) * R - R
laM = np.ceil(laM / R) * R + R
lom = np.floor(lom / R) * R - R
loM = np.ceil(loM / R) * R + R
# %%
lo_range = np.arange(lom, loM + R / 2, R)
la_range = np.arange(lam, laM + R / 2, R)
lo_lab = lo_range[:-1] + R / 2
la_lab = la_range[:-1] + R / 2
# %%
# %%
d, _, _ = np.histogram2d(f['LON'],
f['LAT'],
bins=(lo_range, la_range),
weights=f['HAB'])
from scipy.ndimage import gaussian_filter as gf
d1 = gf(d, sigma=.5)
# %%
ar = xr.DataArray(d1.T,
dims=['LAT', 'LON'],
coords={
'LAT': la_lab,
'LON': lo_lab
})
# %%
de = ar / (R * R * 100 * 100)
# %%
LA = xr.DataArray(f['LAT'], dims=f.index.name)
LO = xr.DataArray(f['LON'], dims=f.index.name)
# %%
de.name = 'DEN'
res = de.interp({'LAT': LA, 'LON': LO}).to_dataframe()
f_out = f.copy()
f_out['DEN'] = res['DEN']
return f_out
def random_abs(absreal,Nrand,wa,er,sl=3,R=20000,ion='HI'):
"""From a real absorber catalog it creates a random catalog. For
a given real absorber with (z_obs,logN_obs,b_obs) it places it at
a new z_rand, defined by where the line could have been
observed.
Input parameters:
---
absreal: numpy rec array with the absorber catalog.
Nrand: number of random lines per real one generated (integer).
wa: numpy array of wavelenght covered by the spectrum.
er: numpy array of error in the normalized flux of the spectrum for
a given wavelenght.
sl: significance level for the detection of the absorption line.
From the error we calculate the Wmin = sl * wa * er / (1+z) / R,
where z = wa/w0 - 1 (w0 is the rest frame wavelenght of the
transition) and R is the resolution of the spectrograp. We then
smooth Wmin with a boxcar (sharp edges).
For the given absorber we transform (logN_obs,b_obs) to a W_obs assuming
linear part of the curve-of-growth.
We then compute the redshifts where W_obs could have been observed
according to the given Wmin, and place Nrand new absorbers with
the same properties as the given one accordingly.
"""
from astro.sampledist import RanDist
from scipy.ndimage import gaussian_filter as gf
absreal.sort(order='LOGN') #np.recarray.sort() sorted by column density
Nrand = int(Nrand)
absrand = absreal.repeat(Nrand)
Ckms = 299792.458
if ion=='HI':
w0 = 1215.67 # HI w0 in angstroms
z = wa/w0 - 1. # spectrum in z coordinates
er = np.where(er==0,1e10,er)
er = np.where(np.isnan(er),1e10,er)
Wmin = 3*sl*w0*er/R #3*sl*wa*er / (1. + z) / R
Wmin = gf(Wmin.astype(float),10) # smoothed version
for i in xrange(len(absreal)):
Wr = logN_b_to_Wr(absreal.LOGN[i],absreal.B[i],ion='HI')
zgood = (Wr > Wmin) & (z>0)
rand_z = RanDist(z, zgood*1.)
zrand = rand_z.random(Nrand)
absrand.ZABS[i*Nrand:(i+1)*Nrand] = zrand
return absrand
def random_abs(absreal, Nrand, wa, er, sl=3, R=20000, ion='HI'):
"""From a real absorber catalog it creates a random catalog. For
a given real absorber with (z_obs,logN_obs,b_obs) it places it at
a new z_rand, defined by where the line could have been
observed.
Input parameters:
---
absreal: numpy rec array with the absorber catalog.
Nrand: number of random lines per real one generated (integer).
wa: numpy array of wavelenght covered by the spectrum.
er: numpy array of error in the normalized flux of the spectrum for
a given wavelenght.
sl: significance level for the detection of the absorption line.
From the error we calculate the Wmin = sl * wa * er / (1+z) / R,
where z = wa/w0 - 1 (w0 is the rest frame wavelenght of the
transition) and R is the resolution of the spectrograp. We then
smooth Wmin with a boxcar (sharp edges).
For the given absorber we transform (logN_obs,b_obs) to a W_obs assuming
linear part of the curve-of-growth.
We then compute the redshifts where W_obs could have been observed
according to the given Wmin, and place Nrand new absorbers with
the same properties as the given one accordingly.
"""
from astro.sampledist import RanDist
from scipy.ndimage import gaussian_filter as gf
absreal.sort(order='LOGN') #np.recarray.sort() sorted by column density
Nrand = int(Nrand)
absrand = absreal.repeat(Nrand)
Ckms = 299792.458
if ion == 'HI':
w0 = 1215.67 # HI w0 in angstroms
z = wa / w0 - 1. # spectrum in z coordinates
er = np.where(er == 0, 1e10, er)
er = np.where(np.isnan(er), 1e10, er)
Wmin = 3 * sl * w0 * er / R #3*sl*wa*er / (1. + z) / R
Wmin = gf(Wmin.astype(float), 10) # smoothed version
for i in xrange(len(absreal)):
Wr = logN_b_to_Wr(absreal.LOGN[i], absreal.B[i], ion='HI')
zgood = (Wr > Wmin) & (z > 0)
rand_z = RanDist(z, zgood * 1.)
zrand = rand_z.random(Nrand)
absrand.ZABS[i * Nrand:(i + 1) * Nrand] = zrand
return absrand
def create_slice():
import numpy as np
from TurbAn.Utilities.subs import create_object
from scipy.ndimage import gaussian_filter as gf
# Create the P3D-Old Object
rc = create_object()
# Ask for variables to extract
vars2ext = input(
"Which variables to extract? e.g. all or bx by bz etc. ").split()
# Set those variables to be loaded in P3D object
rc.vars2load(vars2ext)
# Ask for time slice to read
slice2ext = int(input("Which slice out of " + str(rc.numslices) + "? "))
# Ask if want to smooth data
smooth = int(input("How much smoothing (integer, 0 for none)? "))
if smooth == '': smooth = 0
# Load the time slice
rc.loadslice(slice2ext)
# Write the variables to a file
for i in rc.vars2l:
filename = rc.dirname + "." + i + "." + str(slice2ext) + ".dat"
print(filename)
gf(rc.__dict__[i], sigma=smooth).tofile(filename)
def _smooth(self, xx, yy, H, smooth):
if smooth:
from scipy.ndimage import gaussian_filter as gf
# smooth is assumed to be in velocity units
# So we need to convert it to grid units
# Also note that we are assuming dvx = dvy which does
# not have to be true!
sig = smooth/(xx[1] - xx[0] + yy[1] - yy[0])*2.
#return gf(H, sigma=sig, mode='constant', cval=0.)
return gf(H, sigma=sig)
else:
return H
def __preprocess(self):
logger.info('Pre-processing...')
try:
sss = 0.7
ct = sitk.GetArrayViewFromImage(self.image3D).copy()
ct[ct < -1024] = -1024
sig = min((ct.shape[2] / 150.) * sss, 1)
ct = gf(ct, sig)
ct = self.__calculate_window(ct)
ct = np.array([ct, ct, ct]).transpose(1, 2, 3, 0)
logger.info('Pre-process complete!')
return ct
except Exception as e:
message = (f'Failed to pre-process. Reason: {str(e)}.')
logger.error(message)
def pgmultiplt(rc,variables,bs,fs,step,pgcmp,smooth,numsmooth):
import numpy as np
import pyqtgraph as pg
from pyqtgraph.Qt import QtGui, QtCore
rcd=rc.__dict__
if smooth == 'y':
from scipy.ndimage import gaussian_filter as gf
# Create the window
app=QtGui.QApplication([])
win=pg.GraphicsWindow(title="Multiplot-Test")
win.resize(1000,600)
pg.setConfigOptions(antialias=True)
#Create the canvas
pltdict={}; imgdict={}
for j in rc.vars2l:
idx=rc.vars2l.index(j)
if idx < 4:
haha=[]
elif np.mod(idx,4) == 0:
win.nextRow()
exec('p'+j+'=win.addPlot()')
exec('pltdict["'+j+'"]=p'+j)
if (rc.ny > 1):
exec('img'+j+'=pg.ImageItem()')
exec('imgdict["'+j+'"]=img'+j)
pltdict[j].addItem(imgdict[j])
# pltdict[j].setAspectLocked()
win.show()
# Loop for plottting multiple time slices
for it in range(bs,fs,step):
print('Reading time slice ', it)
rc.loadslice(it)
# Make plots
for j in rc.vars2l:
if (rc.ny==1 and rc.nz==1):
pltdict[j].plot(rc.xx,rcd[j][:,0,0],clear=True)
else:
if smooth == 'y':
imgdict[j].setImage(gf(rcd[j][:,::-1,0].T,sigma=numsmooth),clear=True,lut=pgcmp)
else:
imgdict[j].setImage(rcd[j][:,::-1,0].T,clear=True,lut=pgcmp)
pltdict[j].setTitle(j+' '+sminmax(rcd[j]))
pg.QtGui.QApplication.processEvents()
haha=input('Hello?')
print('All done!')
def psf_recenter(stack, r_mask = 40, cy_ext = 1.5):
'''
find the center of the psf and
stack: the raw_psf
r_mask: the radius of the mask size
cy_ext: how far the background should extend to the outside
'''
nz, ny, nx = stack.shape
cy, cx = np.unravel_index(np.argmax(gf(stack,2)), (nz,ny,nx))[1:]
ny_shift = int(ny/2 - cy)
nx_shift = int(nx/2 - cx)
PSF = np.roll(stack, ny_shift, axis = 1)
PSF = np.roll(PSF, nx_shift, axis = 2)
return PSF
def psf_recenter(stack, r_mask = 40, cy_ext = 1.5):
'''
find the center of the psf and
stack: the raw_psf
r_mask: the radius of the mask size
cy_ext: how far the background should extend to the outside
'''
nz, ny, nx = stack.shape
cy, cx = np.unravel_index(np.argmax(gf(stack,2)), (nz,ny,nx))[1:]
ny_shift = int(ny/2 - cy)
nx_shift = int(nx/2 - cx)
PSF = np.roll(stack, ny_shift, axis = 1)
PSF = np.roll(PSF, nx_shift, axis = 2)
return PSF
def loadslice(self, it, smth=None):
"""
Load the variables initialized by self.vars2load()
"""
if self.data_type in ('b', 'bb'):
for i in self.vars2l:
self.__dict__[i] = self.readslice(self.__dict__[i + 'f'], it,
i)
else:
for i in self.vars2l:
self.__dict__[i] = self.readslice(self.__dict__[i + 'f'], it)
self.mmd = {}
for i in self.vars2l:
if smth is not None:
self.__dict__[i] = gf(self.__dict__[i], sigma=smth)
self.mmd[i] = [self.__dict__[i].min(), self.__dict__[i].max()]
self.time = it * self.dtmovie
def loadslice(self,it,smth=None):
"""
Load the variables initialized by self.vars2load()
"""
for i in self.vars2l:
if i in self.primitives:
self.__dict__[i]=self.readslice(self.__dict__[i+'f'],it,i)
for i in self.vars2l:
if i in self.derived:
#self.__dict__[i] = self._derivedv(i)
self._derivedv(i)
self.mmd={}
for i in self.vars2l:
if smth is not None:
self.__dict__[i]=gf(self.__dict__[i],sigma=smth)
self.mmd[i]=[self.__dict__[i].min(),self.__dict__[i].max()]
self.time = it*self.dtmovie
def xi_gg_LS(self,sigma=0,jacknife=True,f1=None,f2=None,f3=None):
from pyntejos.xcorr.xcorr import W3
from scipy.ndimage import gaussian_filter as gf
s = sigma
W3gg,_ = W3(gf(self.DgDg,s),gf(self.RgRg,s),gf(self.DgRg,s),gf(self.DgRg,s),f1=f1,f2=f2,f3=f3)
#jacknife error
err_W3jk = np.zeros((len(self.rbinedges) - 1, len(self.tbinedges) - 1), float)
if jacknife:
for field in self.fields:
DgDg_aux = self.DgDg - field.DgDg(self.rbinedges,self.tbinedges)
RgRg_aux = self.RgRg - field.RgRg(self.rbinedges,self.tbinedges)
DgRg_aux = self.DgRg - field.DgRg(self.rbinedges,self.tbinedges)
W3gg_aux,_ = W3(gf(DgDg_aux,s),gf(RgRg_aux,s),gf(DgRg_aux,s),gf(DgRg_aux,s),f1=f1,f2=f2,f3=f3)
err_W3jk += (W3gg - W3gg_aux)**2
N = len(self.fields)
err_W3jk = (N - 1.) / N * err_W3jk
err_W3jk = np.sqrt(err_W3jk)
return W3gg, err_W3jk
def __init__(self, cell, ch):
name = cell + '.tif'
movie = Image.open(name)
self.ch = ch[0]
self.sample = ch[1]
self.n_ch = len(ch[0])
self.n_frame = int((movie.n_frames) / self.n_ch)
self.width = movie.width
self.height = movie.height
# I[channel,frame,row,column]
self.I = np.zeros((self.n_ch, self.n_frame, self.height, self.width),
dtype=int)
self.Is = np.zeros((self.n_ch, self.n_frame, self.height, self.width),
dtype=int)
self.mask1 = np.zeros(
(self.n_ch, self.n_frame, self.height, self.width), dtype=bool)
self.mask2 = np.zeros(
(self.n_ch, self.n_frame, self.height, self.width), dtype=bool)
self.Im1 = np.zeros((self.n_ch, self.n_frame, self.height, self.width),
dtype=int)
self.Im2 = np.zeros((self.n_ch, self.n_frame, self.height, self.width),
dtype=int)
for i in range(self.n_ch): # ith channel
movie_i = Image.open(cell + '-' + ch[0][i] + '.tif')
for j in range(self.n_frame): # jth frame
movie_i.seek(j)
I0 = np.array(movie_i, dtype=int)
self.I[i, j] = I0 - I0.min()
self.Is[i, j] = gf(self.I[i, j], sigma)
for j in range(self.n_frame):
for i in range(self.n_ch):
# Mask1 to find a cell from ch2
m1 = self.Is[1, j] > np.percentile(self.Is[1, j], p1)
self.mask1[i, j] = m1
self.Im1[i, j, m1] = self.I[i, j, m1]
self.mask2[i, j, m1] = self.Is[i, j, m1] > np.percentile(
self.Is[i, j, m1], p2)
self.Im2[i, j, self.mask2[i, j]] = self.I[i, j, self.mask2[i,
j]]
def psf_zplane(stack, dz, w0, de = 1):
'''
determine the position of the real focal plane.
Don't mistake with psf_slice!
'''
nz, ny, nx = stack.shape
cy, cx = np.unravel_index(np.argmax(gf(stack,2)), (nz,ny,nx))[1:]
zrange = (nz-1)*dz*0.5
zz = np.linspace(-zrange, zrange, nz)
center_z = stack[:,cy-de:cy+de+1,cx-de:cx+de+1]
im_z = center_z.mean(axis=2).mean(axis=1)
b = np.mean((im_z[0],im_z[-1]))
a = im_z.max() - b
p0 = (a,0,w0,b)
popt = optimize.curve_fit(gaussian, zz, im_z, p0)[0]
z_offset = popt[1] # The original version is wrong
return z_offset, zz
def psf_zplane(stack, dz, w0, de = 1):
'''
determine the position of the real focal plane.
Don't mistake with psf_slice!
'''
nz, ny, nx = stack.shape
cy, cx = np.unravel_index(np.argmax(gf(stack,2)), (nz,ny,nx))[1:]
zrange = (nz-1)*dz*0.5
zz = np.linspace(-zrange, zrange, nz)
center_z = stack[:,cy-de:cy+de+1,cx-de:cx+de+1]
im_z = center_z.mean(axis=2).mean(axis=1)
b = np.mean((im_z[0],im_z[-1]))
a = im_z.max() - b
p0 = (a,0,w0,b)
popt = optimize.curve_fit(gaussian, zz, im_z, p0)[0]
z_offset = popt[1] # The original version is wrong
return z_offset, zz
def binarize(self, sigma=None, smooth=None):
"""
Convert spectra to boolean values at each wavelength.
The procedure estimates the noise by taking the standard deviation
of the derivative spectrum and dividing by sqrt(2). The
zero-point offset for each spectrum is estimated as the mean of
the first 10 wavelengths (empirically seen to be "flat" for most
spectra) and is removed. Resultant points >5sigma [default] are
given a value of True.
Parameters
----------
sigma : float, optional
Sets the threshold, above which the wavelength is considered
to have flux.
"""
# -- smooth if desired
dat = self.data if not smooth else gf(self.data,[smooth,0,0])
if sigma:
# -- estimate the noise and zero point for each spectrum
print("BINARIZE: estimating noise level and zero-point...")
sig = (dat[1:]-dat[:-1])[-100:].std(0)/np.sqrt(2.0)
zer = dat[:10].mean(0)
# -- converting to binary
print("BINARIZE: converting spectra to boolean...")
self.bdata = (dat-zer)>(sigma*sig)
else:
# -- careful about diffraction spikes which look like absoportion
mn_tot = dat.mean(0)
mn_end = dat[-100:].mean(0)
index = mn_tot > mn_end
mn = mn_tot*index + mn_end*~index
# -- binarize by comparison with mean
self.bdata = dat>mn
return
def pderiv(ar,dx=1.,ax=0,order=2,smth=None):
"""
pderiv gives the first partial derivative
of a periodic array along a given axis.
Inputs:
ar - The input array
dx - Grid spacing, defaults to 1.
ax - Axis along which to take the derivative
order - Order of accuracy, (1,2) defaults to 2
Output:
dar - The derivative array
"""
if smth is not None:
ar = gf(ar,sigma=smth)
if order == 1:
dar = (np.roll(ar,-1,axis=ax)-ar)/dx
elif order == 2:
dar = (np.roll(ar,-1,axis=ax)-np.roll(ar,1,axis=ax))/(2*dx)
return dar
def sdss2sl(infile, mask=None, dopcor=False, writetxt=False,
outfile='lixo.txt', gauss_convolve=0, normspec=False,
wlnorm=[6000, 6100], dwl=1, integerwl=True):
"""
Creates an ASCII file from the sdss spectrum file.
Parameters
----------
infile : string
Name of the original SDSS file.
mask : string
Name of the ASCII maks definition file.
dopcor : bool
Apply doppler correction based on the redshift from the
infile header.
write
"""
hdul = pf.open(infile)
wl0 = hdul[0].header['crval1']
npoints = np.shape(hdul[0].data)[1]
dwl = hdul[0].header['cd1_1']
wl = 10 ** (wl0 + np.linspace(0, npoints * dwl, npoints))
if dopcor:
wl = wl / (1. + hdul[0].header['z'])
spectrum = hdul[0].data[0, :] * 1e-17 # in ergs/s/cm^2/A
error = hdul[0].data[2, :] * 1e-17 # in ergs/s/cm^2/A
origmask = hdul[0].data[3, :]
print('Average dispersion: ', np.average(np.diff(wl)))
# Linear interpolation of the spectrum and resampling of
# the spectrum.
f = interp1d(wl, gf(spectrum, gauss_convolve), kind='linear')
err = interp1d(wl, gf(error, gauss_convolve), kind='linear')
om = interp1d(wl, gf(origmask, gauss_convolve), kind='linear')
if integerwl:
wlrebin = np.arange(int(wl[0]) + 1, int(wl[-1]) - 1)
frebin = f(wlrebin)
erebin = err(wlrebin)
mrebin = om(wlrebin)
mcol = np.ones(len(wlrebin))
if mask is not None:
masktab = np.loadtxt(mask)
for i in range(len(masktab)):
mcol[(wlrebin >= masktab[i, 0]) & (wlrebin <= masktab[i, 1])] = 99
else:
mcol = mrebin
mcol[mcol > 3] = 99
vectors = [wlrebin, frebin, erebin, mcol]
txt_format = ['%d', '%.6e', '%.6e', '%d']
slspec = np.column_stack(vectors)
if writetxt:
np.savetxt(outfile, slspec, fmt=txt_format)
return slspec
def fits2sl(spec, mask=None, dwl=1, integerwl=True, writetxt=False,
errfraction=None, normspec=False, normwl=[6100, 6200],
gauss_convolve=0):
"""
Converts a 1D FITS spectrum to the format accepted by Starlight.
Parameters
----------
spec : string
Name of the FITS spectrum.
mask : None or string
Name of the ASCII file containing the regions to be masked,
as a sequence of initial and final wavelength coordinates, one
pair per line.
dwl : number
The step in wavelength of the resampled spectrum. We recommend
using the standard 1 angstrom.
integerwl : boolean
True if the wavelength coordinates can be written as integers.
writetxt : boolean
True if the function should write an output ASCII file.
errfraction : number
Fraction of the signal to be used in case the uncertainties
are unknown.
gauss_convolve : number
Sigma of the gaussian kernel to convolve with the spectrum.
Returns
-------
slspec : numpy.ndarray
2D array with 4 columns, containing wavelength, flux density,
uncertainty and flags respectively.
"""
# Loading spectrum from FITS file.
a = pf.getdata(spec)
wl = get_wl(spec)
print('Average dispersion: ', np.average(np.diff(wl)))
# Linear interpolation of the spectrum and resampling of
# the spectrum.
f = interp1d(wl, gf(a, gauss_convolve), kind='linear')
if integerwl:
wlrebin = np.arange(int(wl[0]) + 1, int(wl[-1]) - 1)
frebin = f(wlrebin)
mcol = np.ones(len(wlrebin))
if mask is not None:
masktab = np.loadtxt(mask)
for i in range(len(masktab)):
mcol[(wlrebin >= masktab[i, 0]) & (wlrebin <= masktab[i, 1])] = 99
if normspec:
normfactor = 1. / np.median(frebin[(wlrebin > normwl[0]) &
(wlrebin < normwl[1])])
else:
normfactor = 1.0
frebin *= normfactor
if (errfraction is not None) and (mask is not None):
vectors = [wlrebin, frebin, frebin * errfraction, mcol]
txt_format = ['%d', '%.6e', '%.6e', '%d']
elif (errfraction is not None) and (mask is None):
vectors = [wlrebin, frebin, frebin * errfraction]
txt_format = ['%d', '%.6e', '%.6e']
elif (errfraction is None) and (mask is not None):
vectors = [wlrebin, frebin, mcol]
txt_format = ['%d', '%.6e', '%d']
elif (errfraction is None) and (mask is None):
vectors = [wlrebin, frebin]
txt_format = ['%d', '%.6e']
slspec = np.column_stack(vectors)
if writetxt:
np.savetxt(spec.strip('fits') + 'txt', slspec, fmt=txt_format)
return slspec
def xradia_star(sh,spokes=48,std=0.5,minfeature=5,ringfact=2,rings=4,contrast=1.,Fast=False):
"""\
creates an Xradia-like star pattern on on array of shape sh
std: "resolution" of xradia star, i.e. standard deviation of the
errorfunction used for smoothing the step (in pixel)
spokes : number of spokes
minfeature : smallest spoke width (in pixel)
ringfact : factorial increase in featuresize from ring to ring.
rings : number of rings
contrast : minimum contrast, set to 0 for gradual color change from zero to 1
set to 1 for no gradient in the spokes
Fast : if set to False, the error function is evaluated at the edges
-> preferred when using fft, as its features are less prone to antiaaliasing
if set to True, simple boolean comparison will be used instead and the
result is later blurred with a gaussian filter.
-> roughly a factor 2 faster
"""
from scipy.ndimage import gaussian_filter as gf
from scipy.special import erf
def step(x,a,std=0.5):
if not Fast:
return 0.5*erf((x-a)/(std*2))+0.5
else:
return (x>a).astype(float)
def rect(x,a):
return step(x,-a/2.,std) * step(-x,-a/2.,std)
def rectint(x,a,b):
return step(x,a,std) * step(-x,-b,std)
ind=np.indices(sh)
cen=(np.array(sh)-1)/2.0
ind=ind-cen.reshape(cen.shape+len(cen)*(1,))
z=ind[1]+1j*ind[0]
spokeint,spokestep=np.linspace(0.0*np.pi,1.0*np.pi,spokes/2,False,True)
spokeint+=spokestep/2
r=np.abs(z)
r0=(minfeature/2.0)/np.sin(spokestep/4.)
rlist=[]
rin=r0
for ii in range(rings):
if rin > max(sh)/np.sqrt(2.):
break
rin*=ringfact
rlist.append((rin*(1-2*np.sin(spokestep/4.)),rin))
spokes=np.zeros(sh)
contrast= np.min((np.abs(contrast),1))
mn=min(spokeint)
mx=max(spokeint)
for a in spokeint:
color = 0.5-np.abs((a-mn)/(mx-mn)-0.5)
spoke=step(np.real(z*np.exp(-1j*(a+spokestep/4))),0)-step(np.real(z*np.exp(-1j*(a-spokestep/4))),0)
spokes+= (spoke*color+0.5*np.abs(spoke))*(1-contrast) + contrast*np.abs(spoke)
spokes*=step(r,r0)
spokes*=step(rlist[-1][0],r)
for ii in range(len(rlist)-1):
a,b=rlist[ii]
spokes*=(1.0-rectint(r,a,b))
if Fast:
return gf(spokes,std)
else:
return spokes
def acquirePSF(self, range_, nSlices, nFrames, center_xy=True, filename=None,
mask_size = 100, mask_center = (-1,-1)):
'''
Acquires a PSF stack. The PSF is returned but also stored internally.
:Parameters:
*range_*: float
The range of the scan around the current axial position in micrometers.
*nSlices*: int
The number of PSF slices to acquire.
*nFrames*: int
The number of frames to be averaged for each PSF slice.
*filename*: str
The file name into which the PSF will be saved.
:Returns:
*PSF*: numpy.array
An array of shape (k,l,m), where k are the number of PSF slices and
(l,m) the lateral slice dimensions.
'''
# Logging
self._settings['range'] = range_
self._settings['nSlices'] = nSlices
self._settings['nFrames'] = nFrames
self._settings['filename'] = filename
# Some parameters
start = range_/2.0
end = -range_/2.0
self._dz = abs(range_/(nSlices-1.0))
# Scan the PSF:
scan = self._control.piezoscan.scan(start, end, nSlices, nFrames, filename)
nz, nx, ny = scan.shape
# An empty PSF
PSF = np.zeros_like(scan)
# Geometry info
g = pupil.Geometry((nx,ny), nx/2.-0.5, ny/2.-0.5, 16)
# Filled cylinder:
cyl = np.array(nz*[g.r_pxl<16])
new_cyl = np.array(nz*[g.r_pxl<mask_size])
# Fill the empty PSF with every voxel in scan where cyl=True
if not center_xy:
PSF[cyl] = scan[cyl]
# Hollow cylinder
hcyl = np.array(nz*[np.logical_and(g.r_pxl>=50, g.r_pxl<61)])
if mask_center[0] > -1:
g2 = pupil.Geometry((nx,ny), mask_center[0], mask_center[1], 16)
mask = g2.r_pxl<mask_size
else:
mask = g.r_pxl<mask_size
if filename:
np.save(filename+"pre-cut", scan)
if center_xy:
# The coordinates of the brightest pixel
cz, cx, cy = np.unravel_index(np.argmax(gf(scan*mask,2)), scan.shape)
print "Center found at: ", (cz,cx,cy)
# We laterally center the scan at the brightest pixel
cut = scan[:,cx-mask_size:cx+mask_size,cy-mask_size:cy+mask_size]
PSF = np.zeros((nz,nx,ny))
PSF[:,nx/2-mask_size:nx/2+mask_size,ny/2-mask_size:ny/2+mask_size] = cut
# Background estimation
self._background = np.mean(scan[hcyl])
print "Background guess: ", self._background
else:
self._background = np.mean(scan[hcyl])
PSF[np.logical_not(new_cyl)] = self._background
self._PSF = PSF
if filename:
np.save(filename, PSF)
return PSF
