def __init__(self, cube, beam=None, mask=None, method="MAD", compute=True):
# Initialize cube object
self.cube = SpectralCube.read(cube)
if mask is not None:
_check_mask(mask)
self.mask = mask
if beam is not None:
_check_beam(mask)
# Initialize noise object
self.noise = Noise(self.cube, beam=beam, method=method)
def moment_masking(cube, kernel_size, clip=5, dilations=1):
'''
'''
if not signal_id_flag:
raise ImportError("signal-id is not installed."
" This function is not available.")
smooth_data = convolve(cube.filled_data[:], gauss_kern(kernel_size))
fake_mask = LazyMask(np.isfinite, cube=cube)
smooth_cube = SpectralCube(data=smooth_data, wcs=cube.wcs, mask=fake_mask)
smooth_scale = Noise(smooth_cube).scale
mask = (smooth_cube > (clip * smooth_scale)).include()
# Now dilate the mask once
dilate_struct = nd.generate_binary_structure(3, 3)
mask = nd.binary_dilation(mask,
structure=dilate_struct,
iterations=dilations)
return mask
def _update(self, data=None, wcs=None, beam=None, method="MAD"):
'''
Helper function to update classes.
'''
# Check if we need a new SpectralCube
if data is None and wcs is None:
pass
else:
if data is None:
data = self.cube.unmasked_data[:]
if wcs is None:
wcs = self.cube.wcs
# Make new SpectralCube object
self.cube = SpectralCube(data=data, wcs=wcs)
if beam is not None:
_check_beam(beam)
self.noise = Noise(self.cube, beam=beam, method=method)
def __init__(self,
cube,
beam=None,
mask=None,
method="MAD",
compute=True):
raise NotImplementedError("SimCube is not yet implemented for "
"general use.")
# Initialize cube object
self.cube = SpectralCube.read(cube)
if mask is not None:
_check_mask(mask)
self.mask = mask
if beam is not None:
_check_beam(mask)
# Initialize noise object
self.noise = Noise(self.cube, beam=beam, method=method)
def __init__(self, cube, beam=None, mask=None, method="MAD", compute=True):
# Initialize cube object
self.cube = SpectralCube.read(cube)
if mask is not None:
_check_mask(mask)
self.mask = mask
if beam is not None:
_check_beam(mask)
# Initialize noise object
self.noise = Noise(self.cube, beam=beam, method=method)
def __init__(self, cube, mask=None, algorithm=None, beam=None):
super(ObsCube, self).__init__()
self.cube = sc.SpectralCube.read(cube)
self.algorithm = algorithm
# Make sure mask is an accepted type
if mask is not None:
_check_mask(mask)
self.mask = mask
if beam is not None:
_check_beam(beam)
self.noise = Noise(self.cube, beam=beam)
def __init__(self,
cube,
noise_type='constant',
clip=3,
scale=None,
moment_method='slice'):
super(Mask_and_Moments, self).__init__()
if not spectral_cube_flag:
raise ImportError("Mask_and_Moments requires the spectral-cube "
" to be installed: https://github.com/"
"radio-astro-tools/spectral-cube")
if isinstance(cube, SpectralCube):
self.cube = cube
self.save_name = None
else:
self.cube = SpectralCube.read(cube)
# Default save name to the cube name without the suffix.
self.save_name = ".".join(cube.split(".")[:-1])
self.noise_type = noise_type
self.clip = clip
if moment_method not in ['slice', 'cube', 'ray']:
raise TypeError("Moment method must be 'slice', 'cube', or 'ray'.")
self.moment_how = moment_method
if scale is None:
if not signal_id_flag:
raise ImportError("signal-id is not installed and error"
" estimation is not available. You must "
"provide the noise scale.")
self.scale = Noise(self.cube).scale * self.cube.unit
else:
if not isinstance(scale, u.Quantity):
raise TypeError("scale must be a Quantity with the same units"
" as the given cube.")
if not scale.unit == self.cube.unit:
raise u.UnitsError("scale must have the same units"
" as the given cube.")
self.scale = scale
self.prop_headers = None
self.prop_err_headers = None
def __init__(self, cube, mask=None, algorithm=None, beam=None):
raise NotImplementedError("ObsCube is not yet implemented for "
"general use.")
super(ObsCube, self).__init__()
self.cube = SpectralCube.read(cube)
self.algorithm = algorithm
# Make sure mask is an accepted type
if mask is not None:
_check_mask(mask)
self.mask = mask
if beam is not None:
_check_beam(beam)
self.noise = Noise(self.cube, beam=beam)
def _update(self, data=None, wcs=None, beam=None, method="MAD"):
'''
Helper function to update classes.
'''
# Check if we need a new SpectralCube
if data is None and wcs is None:
pass
else:
if data is None:
data = self.cube.unmasked_data[:]
if wcs is None:
wcs = self.cube.wcs
# Make new SpectralCube object
self.cube = SpectralCube(data=data, wcs=wcs)
if beam is not None:
_check_beam(beam)
self.noise = Noise(self.cube, beam=beam, method=method)
def find_noise(self, return_obj=False):
'''
Returns noise estimate, or the whole Noise object.
Parameters
----------
return_obj : bool, optional
If True, returns the Noise object. Otherwise returns the estimated
noise level.
'''
if not signal_id_flag:
raise ImportError("signal-id is not installed."
" This function is not available.")
noise = Noise(self.cube)
self.scale = noise.scale
if return_obj:
return noise
return noise.scale
smooth_chans = int(round_up_to_odd(200. / 66.))
# 10 consecutive channels must be above the MAD level to be real emission.
num_chans = 7
peak_snr = 4.5
# Cutoff level
min_snr = 3
# Where to cut at the line edges
edge_thresh = 0.5
# Smooth the cube, then create a noise model
spec_kernel = Box1DKernel(smooth_chans)
smooth_cube = cube.spectral_smooth(spec_kernel)
noise = Noise(smooth_cube)
noise.estimate_noise(spectral_flat=True)
noise.get_scale_cube()
snr = noise.snr.copy()
posns = np.where(snr.max(axis=0) >= min_snr)
bad_pos = np.where(snr.max(axis=0) < min_snr)
mask[:, bad_pos[0], bad_pos[1]] = False
# In case single spectra need to be inspected.
verbose = False
for i, j in ProgressBar(zip(*posns)):
smooth_chans = int(round_up_to_odd(200. / 66.))
# 10 consecutive channels must be above the MAD level to be real emission.
num_chans = 7
peak_snr = 4.5
# Cutoff level
min_snr = 3
# Where to cut at the line edges
edge_thresh = 0.5
# Smooth the cube, then create a noise model
spec_kernel = Box1DKernel(smooth_chans)
smooth_cube = cube.spectral_smooth(spec_kernel)
noise = Noise(smooth_cube)
noise.estimate_noise(spectral_flat=True)
noise.get_scale_cube()
snr = noise.snr.copy()
posns = np.where(snr.max(axis=0) >= min_snr)
bad_pos = np.where(snr.max(axis=0) < min_snr)
mask[:, bad_pos[0], bad_pos[1]] = False
# In case single spectra need to be inspected.
verbose = False
for i, j in ProgressBar(zip(*posns)):
def make_signal_mask(cube, smooth_chans=200. / 66., min_chan=7, peak_snr=5.,
min_snr=3.5, edge_thresh=1.5, verbose=False):
'''
Create a robust signal mask by requiring spatial and spectral
connectivity.
'''
import astropy.units as u
from astropy.convolution import Box1DKernel
from signal_id import Noise
from scipy import ndimage as nd
from astropy.wcs.utils import proj_plane_pixel_scales
from astropy.utils.console import ProgressBar
import skimage.morphology as mo
import numpy as np
from radio_beam import Beam
from itertools import groupby, chain
from operator import itemgetter
import matplotlib.pyplot as p
pixscale = proj_plane_pixel_scales(cube.wcs)[0]
# # Want to smooth the mask edges
mask = cube.mask.include().copy()
# Set smoothing parameters and # consecutive channels.
smooth_chans = int(round_up_to_odd(smooth_chans))
# consecutive channels to be real emission.
num_chans = min_chan
# Smooth the cube, then create a noise model
spec_kernel = Box1DKernel(smooth_chans)
smooth_cube = cube.spectral_smooth(spec_kernel)
noise = Noise(smooth_cube)
noise.estimate_noise(spectral_flat=True)
noise.get_scale_cube()
snr = noise.snr.copy()
snr[np.isnan(snr)] = 0.0
posns = np.where(snr.max(axis=0) >= min_snr)
bad_pos = np.where(snr.max(axis=0) < min_snr)
mask[:, bad_pos[0], bad_pos[1]] = False
for i, j in ProgressBar(zip(*posns)):
# Look for all pixels above min_snr
good_posns = np.where(snr[:, i, j] > min_snr)[0]
# Reject if the total is less than connectivity requirement
if good_posns.size < num_chans:
mask[:, i, j] = False
continue
# Find connected pixels
sequences = []
for k, g in groupby(enumerate(good_posns), lambda (i, x): i - x):
sequences.append(map(itemgetter(1), g))
# Check length and peak. Require a minimum of 3 pixels above the noise
# to grow from.
sequences = [seq for seq in sequences if len(seq) >= 3 and
np.nanmax(snr[:, i, j][seq]) >= peak_snr]
# Continue if no good sequences found
if len(sequences) == 0:
mask[:, i, j] = False
continue
# Now take each valid sequence and expand the edges until the smoothed
# spectrum approaches zero.
edges = [[seq[0], seq[-1]] for seq in sequences]
for n, edge in enumerate(edges):
# Lower side
if n == 0:
start_posn = edge[0]
stop_posn = 0
else:
start_posn = edge[0] - edges[n - 1][0]
stop_posn = edges[n - 1][0]
for pt in np.arange(start_posn, stop_posn, -1):
# if smoothed[pt] <= mad * edge_thresh:
if snr[:, i, j][pt] <= edge_thresh:
break
sequences[n].insert(0, pt)
# Upper side
start_posn = edge[1]
if n == len(edges) - 1:
stop_posn = cube.shape[0]
else:
stop_posn = edges[n + 1][0]
for pt in np.arange(start_posn, stop_posn, 1):
# if smoothed[pt] <= mad * edge_thresh:
if snr[:, i, j][pt] <= edge_thresh:
break
sequences[n].insert(0, pt)
# Final check for the min peak level and ensure all meet the
# spectral connectivity requirement
sequences = [seq for seq in sequences if len(seq) >= num_chans and
np.nanmax(snr[:, i, j][seq]) >= peak_snr]
if len(sequences) == 0:
mask[:, i, j] = False
continue
bad_posns = \
list(set(np.arange(cube.shape[0])) - set(list(chain(*sequences))))
mask[:, i, j][bad_posns] = False
if verbose:
p.subplot(121)
p.plot(cube.spectral_axis.value, noise.snr[:, i, j])
min_val = cube.spectral_axis.value[np.where(mask[:, i, j])[0][-1]]
max_val = cube.spectral_axis.value[np.where(mask[:, i, j])[0][0]]
p.vlines(min_val, 0,
np.nanmax(noise.snr[:, i, j]))
p.vlines(max_val, 0,
np.nanmax(noise.snr[:, i, j]))
p.plot(cube.spectral_axis.value,
noise.snr[:, i, j] * mask[:, i, j], 'bD')
p.subplot(122)
p.plot(cube.spectral_axis.value, cube[:, i, j], label='Cube')
p.plot(cube.spectral_axis.value, smooth_cube[:, i, j],
label='Smooth Cube')
p.axvline(min_val)
p.axvline(max_val)
p.plot(cube.spectral_axis.value,
smooth_cube[:, i, j] * mask[:, i, j], 'bD')
p.draw()
raw_input("Next spectrum?")
p.clf()
sub1.show_colorbar()
sub2.show_colorbar()
sub1.colorbar.set_axis_label_text("Intensity (Jy/beam)")
sub2.hide_yaxis_label()
chan_fig.tight_layout()
raw_input("Continue?")
chan_fig.close()
# Now create good masks and derive 0th moments.
noise_cube = Noise(cube)
new_noise = noise_cube.scale
cube = cube.with_mask(cube > new_noise*u.Jy)
old_noise_cube = Noise(old_cube)
old_noise = old_noise_cube.scale
old_cube = old_cube.with_mask(old_cube > old_noise*u.Jy)
# Load in the broad clean mask used
clean_mask = fits.getdata("../../../Arecibo/M33_newmask.fits")
# Need to match the dims
class SimCube(object):
'''
A wrapping class to prepare a simulated spectral data cube for
comparison with another cube.
'''
def __init__(self, cube, beam=None, mask=None, method="MAD", compute=True):
# Initialize cube object
self.cube = SpectralCube.read(cube)
if mask is not None:
_check_mask(mask)
self.mask = mask
if beam is not None:
_check_beam(mask)
# Initialize noise object
self.noise = Noise(self.cube, beam=beam, method=method)
def add_noise(self):
'''
Use Noise to add synthetic noise to the data. Then update
SpectralCube.
'''
# Create the noisy cube
self.noise.get_noise_cube()
noise_data = self.noise.noise_cube +\
self.cube.filled_data[:]
# Update SpectralCube object
self._update(data=noise_data)
return self
def clean_cube(self, algorithm=None):
raise NotImplementedError("")
def apply_mask(self, mask=None):
'''
Check if the given mask is acceptable abd apply to
SpectralCube.
'''
# Update mask
if mask is not None:
_check_mask(mask)
self.mask = mask
# Create the mask, auto masking nan values
default_mask = np.isfinite(self.cube.filled_data[:])
if self.mask is not None:
self.mask = CompositeMask(default_mask, self.mask)
else:
self.mask = default_mask
# Apply mask to spectral cube object
self.cube = self.cube.with_mask(mask)
return self
def _update(self, data=None, wcs=None, beam=None, method="MAD"):
'''
Helper function to update classes.
'''
# Check if we need a new SpectralCube
if data is None and wcs is None:
pass
else:
if data is None:
data = self.cube.unmasked_data[:]
if wcs is None:
wcs = self.cube.wcs
# Make new SpectralCube object
self.cube = SpectralCube(data=data, wcs=wcs)
if beam is not None:
_check_beam(beam)
self.noise = Noise(self.cube, beam=beam, method=method)
def compute_properties(self):
'''
Use SpectralCube to compute the moments. Also compute the integrated
intensity based on the noise properties from Noise.
'''
self._moment0 = self.cube.moment0().value
self._moment1 = self.cube.moment1().value
self._moment2 = self.cube.moment2().value
_get_int_intensity(self)
return self
@property
def moment0(self):
return self._moment0
@property
def moment1(self):
return self._moment1
@property
def moment2(self):
return self._moment2
@property
def intint(self):
return self._intint
def sim_prep(self, mask=None):
'''
Prepares the cube when being compared to another simulation.
This entails:
* Optionally applying a mask to the data.
* Computing the cube's property arrays
'''
if not mask is None:
self.apply_mask()
self.compute_properties()
return self
def obs_prep(self, mask=None):
'''
Prepares the cube when being compared to observational data.
This entails:
* Optionally applying a mask to the data.
* Add synthetic noise based on the cube's properties.
* Computing the cube's property arrays
'''
if not mask is None:
self.apply_mask()
self.add_noise()
self.compute_properties()
return self
class SimCube(object):
'''
A wrapping class to prepare a simulated spectral data cube for
comparison with another cube.
'''
def __init__(self, cube, beam=None, mask=None, method="MAD", compute=True):
# Initialize cube object
self.cube = SpectralCube.read(cube)
if mask is not None:
_check_mask(mask)
self.mask = mask
if beam is not None:
_check_beam(mask)
# Initialize noise object
self.noise = Noise(self.cube, beam=beam, method=method)
def add_noise(self):
'''
Use Noise to add synthetic noise to the data. Then update
SpectralCube.
'''
# Create the noisy cube
self.noise.get_noise_cube()
noise_data = self.noise.noise_cube +\
self.cube.filled_data[:]
# Update SpectralCube object
self._update(data=noise_data)
return self
def clean_cube(self, algorithm=None):
raise NotImplementedError("")
def apply_mask(self, mask=None):
'''
Check if the given mask is acceptable abd apply to
SpectralCube.
'''
# Update mask
if mask is not None:
_check_mask(mask)
self.mask = mask
# Create the mask, auto masking nan values
default_mask = np.isfinite(self.cube.filled_data[:])
if self.mask is not None:
self.mask = CompositeMask(default_mask, self.mask)
else:
self.mask = default_mask
# Apply mask to spectral cube object
self.cube = self.cube.with_mask(mask)
return self
def _update(self, data=None, wcs=None, beam=None, method="MAD"):
'''
Helper function to update classes.
'''
# Check if we need a new SpectralCube
if data is None and wcs is None:
pass
else:
if data is None:
data = self.cube.unmasked_data[:]
if wcs is None:
wcs = self.cube.wcs
# Make new SpectralCube object
self.cube = SpectralCube(data=data, wcs=wcs)
if beam is not None:
_check_beam(beam)
self.noise = Noise(self.cube, beam=beam, method=method)
def compute_properties(self):
'''
Use SpectralCube to compute the moments. Also compute the integrated
intensity based on the noise properties from Noise.
'''
self._moment0 = self.cube.moment0().value
self._moment1 = self.cube.moment1().value
self._moment2 = self.cube.moment2().value
_get_int_intensity(self)
return self
@property
def moment0(self):
return self._moment0
@property
def moment1(self):
return self._moment1
@property
def moment2(self):
return self._moment2
@property
def intint(self):
return self._intint
def sim_prep(self, mask=None):
'''
Prepares the cube when being compared to another simulation.
This entails:
* Optionally applying a mask to the data.
* Computing the cube's property arrays
'''
if not mask is None:
self.apply_mask()
self.compute_properties()
return self
def obs_prep(self, mask=None):
'''
Prepares the cube when being compared to observational data.
This entails:
* Optionally applying a mask to the data.
* Add synthetic noise based on the cube's properties.
* Computing the cube's property arrays
'''
if not mask is None:
self.apply_mask()
self.add_noise()
self.compute_properties()
return self