def vis_wstack_iter(vis, **kwargs):
""" W slice iterator
:param wstack: wstack (wavelengths)
:param vis_slices: Number of slices (second in precedence to wstack)
:returns: Boolean array with selected rows=True
"""
assert type(vis) == Visibility or type(vis) == BlockVisibility
wmaxabs = (numpy.max(numpy.abs(vis.w)))
wstack = get_parameter(kwargs, "wstack", None)
if wstack is None:
vis_slices = get_parameter(kwargs, "vis_slices", 1)
boxes = numpy.linspace(-wmaxabs, +wmaxabs, vis_slices)
wstack = 2 * wmaxabs / vis_slices
else:
vis_slices = 1 + 2 * numpy.round(wmaxabs / wstack).astype('int')
boxes = numpy.linspace(-wmaxabs, +wmaxabs, vis_slices)
for box in boxes:
rows = numpy.abs(vis.w - box) < 0.5 * wstack
if numpy.sum(rows) > 0:
yield rows
else:
yield None
def vis_timeslice_iter(vis: Visibility, **kwargs) -> numpy.ndarray:
""" W slice iterator
:param wstack: wstack (wavelengths)
:param vis_slices: Number of slices (second in precedence to wstack)
:return: Boolean array with selected rows=True
"""
assert isinstance(vis, Visibility) or isinstance(vis, BlockVisibility), vis
timemin = numpy.min(vis.time)
timemax = numpy.max(vis.time)
timeslice = get_parameter(kwargs, "timeslice", 'auto')
if timeslice == 'auto':
boxes = numpy.unique(vis.time)
timeslice = 0.1
elif timeslice is None:
timeslice = timemax - timemin
boxes = [0.5 * (timemax + timemin)]
elif isinstance(timeslice, float) or isinstance(timeslice, int):
boxes = numpy.arange(timemin, timemax, timeslice)
else:
vis_slices = get_parameter(kwargs, "vis_slices", None)
assert vis_slices is not None, "Time slicing not specified: set either timeslice or vis_slices"
boxes = numpy.linspace(timemin, timemax, vis_slices)
if vis_slices > 1:
timeslice = boxes[1] - boxes[0]
else:
timeslice = timemax - timemin
for box in boxes:
rows = numpy.abs(vis.time - box) <= 0.5 * timeslice
yield rows
def add_noise_to_visibility(vis, polarisation='stokesI', **kwargs):
""" Add noise to visibility from the 'standard normal' distribution, but optionally the distribution mu and sigma can be changed
The function should be modified for different polarization setupd (e.g stokesIQ)
:param vis:
:param polarisation: the polarizations which the noise need to be added (currently works only for stokes polarization frame)
:return:
"""
sigma = get_parameter(kwargs, "sigma", None);#tuple of the sigma of the real and the imaginary part
if sigma is None:
sigma = (1,1);
mu = get_parameter(kwargs, "mu", None);#tuple of the mu of the real and the imaginary part
if mu is None:
mu = (0,0);
log.info("add_noise_to_visibility: Add noise to visibilities using the normal distribution (with mu=%.2f, %.2f sigma=%.2f, %.2f to Re and Im respectively)"
%(mu[0],mu[1],sigma[0],sigma[1]));
num_of_visibilities = vis.data['vis'].shape[0];
if polarisation == 'stokesI':
### no need to check polarisation frame as it is working with stokesI and stokesIQUV as well ###
#assert vis.polarisation_frame == PolarisationFrame("stokesI");
vis.data['vis'][:,0] += numpy.vectorize(complex)(sigma[0] * numpy.random.randn(num_of_visibilities) + mu[0]
,sigma[1] * numpy.random.randn(num_of_visibilities) + mu[1]);
if polarisation == 'stokesIQUV':
assert vis.polarisation_frame == PolarisationFrame("stokesIQUV");
for i in range(0,4):
vis.data['vis'][:,i] += numpy.vectorize(complex)(sigma[0] * numpy.random.randn(num_of_visibilities) + mu[0]
,sigma[1] * numpy.random.randn(num_of_visibilities) + mu[1]);
return vis;
def vis_timeslice_iter(vis: Visibility, **kwargs) -> numpy.ndarray:
""" W slice iterator
:param wstack: wstack (wavelengths)
:param vis_slices: Number of slices (second in precedence to wstack)
:return: Boolean array with selected rows=True
"""
assert isinstance(vis, Visibility) or isinstance(vis, BlockVisibility)
timemin = numpy.min(vis.time)
timemax = numpy.max(vis.time)
timeslice = get_parameter(kwargs, "timeslice", None)
if timeslice is None or timeslice == 'auto':
vis_slices = get_parameter(kwargs, "vis_slices", None)
if vis_slices is None or vis_slices == 'auto':
vis_slices = len(numpy.unique(vis.time))
boxes = numpy.linspace(timemin, timemax, vis_slices)
timeslice = (timemax - timemin) / vis_slices
else:
vis_slices = 1 + 2 * numpy.ceil(
(timemax - timemin) / timeslice).astype('int')
boxes = numpy.linspace(timemin, timemax, vis_slices)
if vis_slices > 1:
timeslice = boxes[1] - boxes[0]
else:
timeslice = timemax - timemin
for box in boxes:
rows = numpy.abs(vis.time - box) <= 0.5 * timeslice
yield rows
def vis_wstack_iter(vis: Visibility, **kwargs) -> numpy.ndarray:
""" W slice iterator
:param wstack: wstack (wavelengths)
:param vis_slices: Number of slices (second in precedence to wstack)
:return: Boolean array with selected rows=True
"""
assert isinstance(vis, Visibility) or isinstance(vis, BlockVisibility)
wmaxabs = numpy.max(numpy.abs(vis.w))
wstack = get_parameter(kwargs, "wstack", None)
if wstack is None:
vis_slices = get_parameter(kwargs, "vis_slices", 1)
boxes = numpy.linspace(-wmaxabs, wmaxabs, vis_slices)
if vis_slices > 1:
wstack = boxes[1] - boxes[0]
else:
wstack = 2 * wmaxabs
else:
vis_slices = 1 + 2 * numpy.round(wmaxabs / wstack).astype('int')
boxes = numpy.linspace(-wmaxabs, +wmaxabs, vis_slices)
if vis_slices > 1:
wstack = boxes[1] - boxes[0]
else:
wstack = 2 * wmaxabs
for box in boxes:
rows = numpy.abs(vis.w - box) < 0.5 * wstack
yield rows
def create_predict_gleam_model_graph(vis_graph_list,
frequency,
channel_bandwidth,
npixel=512,
cellsize=0.001,
**kwargs):
""" Create a graph to fill in a model with the gleam sources and predict into a vis_graph_list
:param vis_graph_list:
:param frequency:
:param channel_bandwidth:
:param npixel: 512
:param cellsize: 0.001
:param kwargs:
:return: vis_graph_list
"""
# Note that each vis_graph has it's own model_graph
predicted_vis_graph_list = list()
for i, vis_graph in enumerate(vis_graph_list):
facets = {}
if get_parameter(kwargs, "facets", False):
facets = {'facets': get_parameter(kwargs, "facets", False)}
model_graph = delayed(create_low_test_image_from_gleam)(
vis_graph,
frequency,
channel_bandwidth,
npixel=npixel,
cellsize=cellsize,
**facets)
predicted_vis_graph_list.append(
create_predict_graph([vis_graph], model_graph, **kwargs)[0])
return predicted_vis_graph_list
def weight_visibility(vis: Visibility, im: Image, **kwargs) -> Visibility:
""" Reweight the visibility data using a selected algorithm
Imaging uses the column "imaging_weight" when imaging. This function sets that column using a
variety of algorithms
Options are:
- Natural: by visibility weight (optimum for noise in final image)
- Uniform: weight of sample divided by sum of weights in cell (optimum for sidelobes)
- Super-uniform: As uniform, by sum of weights is over extended box region
- Briggs: Compromise between natural and uniform
- Super-briggs: As Briggs, by sum of weights is over extended box region
:param vis:
:param im:
:return: visibility with imaging_weights column added and filled
"""
assert isinstance(vis, Visibility), "vis is not a Visibility: %r" % vis
assert get_parameter(kwargs, "padding", False) is False
spectral_mode, vfrequencymap = get_frequency_map(vis, im)
polarisation_mode, vpolarisationmap = get_polarisation_map(vis, im)
uvw_mode, shape, padding, vuvwmap = get_uvw_map(vis, im)
density = None
densitygrid = None
weighting = get_parameter(kwargs, "weighting", "uniform")
vis.data['imaging_weight'], density, densitygrid = weight_gridding(
im.data.shape, vis.data['weight'], vuvwmap, vfrequencymap,
vpolarisationmap, weighting)
return vis, density, densitygrid
def gaintable_timeslice_iter(gt: GainTable, **kwargs) -> numpy.ndarray:
""" W slice iterator
:param wstack: wstack (wavelengths)
:param gt_slices: Number of slices (second in precedence to wstack)
:return: Boolean array with selected rows=True
"""
assert isinstance(gt, GainTable)
timemin = numpy.min(gt.time)
timemax = numpy.max(gt.time)
timeslice = get_parameter(kwargs, "timeslice", 'auto')
if timeslice == 'auto':
boxes = numpy.unique(gt.time)
timeslice = 0.1
elif timeslice is None:
timeslice = timemax - timemin
boxes = [0.5 * (timemax + timemin)]
elif isinstance(timeslice, float) or isinstance(timeslice, int):
boxes = numpy.arange(timemin, timemax, timeslice)
else:
gt_slices = get_parameter(kwargs, "gaintable_slices", None)
assert gt_slices is not None, "Time slicing not specified: set either timeslice or gt_slices"
boxes = numpy.linspace(timemin, timemax, gt_slices)
if gt_slices > 1:
timeslice = boxes[1] - boxes[0]
else:
timeslice = timemax - timemin
for box in boxes:
rows = numpy.abs(gt.time - box) <= 0.5 * timeslice
yield rows
def invert_2d(vis: Visibility,
im: Image,
dopsf=False,
normalize=True,
**kwargs) -> (Image, numpy.ndarray):
""" Invert using prolate spheroidal gridding function
Use the image im as a template. Do PSF in a separate call.
Note that the image is not normalised but the sum of the weights. This is for ease of use in partitioning.
:param vis: Visibility to be inverted
:param im: image template (not changed)
:param dopsf: Make the psf instead of the dirty image
:param normalize: Normalize by the sum of weights (True)
:return: resulting image[nchan, npol, ny, nx], sum of weights[nchan, npol]
"""
log.debug("invert_2d: inverting using 2d transform")
kwargs['kernel'] = get_parameter(kwargs, "kernel", '2d')
timing = get_parameter(kwargs, "timing", False)
if timing:
return invert_2d_base_timing(vis,
im,
dopsf,
normalize=normalize,
**kwargs)
return invert_2d_base(vis, im, dopsf, normalize=normalize, **kwargs)
def coalesce_visibility(vis: BlockVisibility, **kwargs) -> Visibility:
""" Coalesce the BlockVisibility data. The output format is a Visibility, as needed for imaging
Coalesce by baseline-dependent averaging (optional). The number of integrations averaged goes as the ratio of the
maximum possible baseline length to that for this baseline. This number can be scaled by coalescence_factor and
limited by max_coalescence.
When faceting, the coalescence factors should be roughly the same as the number of facets on one axis.
If coalescence_factor=0.0 then just a format conversion is done
:param vis: BlockVisibility to be coalesced
:return: Coalesced visibility with cindex and blockvis filled in
"""
# 将blockvisibility合并为一个visibility
assert type(
vis) is BlockVisibility, "vis is not a BlockVisibility: %r" % vis
time_coal = get_parameter(kwargs, 'time_coal', 0.0)
max_time_coal = get_parameter(kwargs, 'max_time_coal', 100)
frequency_coal = get_parameter(kwargs, 'frequency_coal', 0.0)
max_frequency_coal = get_parameter(kwargs, 'max_frequency_coal', 100)
if time_coal == 0.0 and frequency_coal == 0.0:
# 若time和frequency的合并率均为0，则只做一个简单的blockvisibility到visibility的变换
return convert_blockvisibility_to_visibility((vis))
# 否则使用average_in_blocks做一系列较为复杂的合并变换
cvis, cuvw, cwts, ctime, cfrequency, cchannel_bandwidth, ca1, ca2, cintegration_time, cindex \
= average_in_blocks(vis.data['vis'], vis.data['uvw'], vis.data['weight'], vis.time, vis.integration_time,
vis.frequency, vis.channel_bandwidth, time_coal, max_time_coal,
frequency_coal, max_frequency_coal)
cimwt = numpy.ones(cvis.shape)
coalesced_vis = Visibility(uvw=cuvw,
time=ctime,
frequency=cfrequency,
channel_bandwidth=cchannel_bandwidth,
phasecentre=vis.phasecentre,
antenna1=ca1,
antenna2=ca2,
vis=cvis,
weight=cwts,
imaging_weight=cimwt,
configuration=vis.configuration,
integration_time=cintegration_time,
polarisation_frame=vis.polarisation_frame,
cindex=cindex,
blockvis=vis)
log.debug(
'coalesce_visibility: Created new Visibility for coalesced data, coalescence factors (t,f) = (%.3f,%.3f)'
% (time_coal, frequency_coal))
log.debug('coalesce_visibility: Maximum coalescence (t,f) = (%d, %d)' %
(max_time_coal, max_frequency_coal))
log.debug('coalesce_visibility: Original %s, coalesced %s' %
(vis_summary(vis), vis_summary(coalesced_vis)))
return coalesced_vis
def solve_image(vis: Visibility,
model: Image,
components=None,
predict=predict_2d,
invert=invert_2d,
**kwargs) -> (Visibility, Image, Image):
"""Solve for image using deconvolve_cube and specified predict, invert
This is the same as a majorcycle/minorcycle algorithm. The components are removed prior to deconvolution.
See also arguments for predict, invert, deconvolve_cube functions.2d
:param vis:
:param model: Model image
:param predict: Predict function e.g. predict_2d, predict_wstack
:param invert: Invert function e.g. invert_2d, invert_wstack
:return: Visibility, model
"""
nmajor = get_parameter(kwargs, 'nmajor', 5)
log.info("solve_image: Performing %d major cycles" % nmajor)
# The model is added to each major cycle and then the visibilities are
# calculated from the full model
vispred = copy_visibility(vis)
visres = copy_visibility(vis)
vispred = predict(vispred, model, **kwargs)
if components is not None:
vispred = predict_skycomponent_visibility(vispred, components)
visres.data['vis'] = vis.data['vis'] - vispred.data['vis']
dirty, sumwt = invert(visres, model, **kwargs)
psf, sumwt = invert(visres, model, dopsf=True, **kwargs)
thresh = get_parameter(kwargs, "threshold", 0.0)
for i in range(nmajor):
log.info("solve_image: Start of major cycle %d" % i)
cc, res = deconvolve_cube(dirty, psf, **kwargs)
res = None
model.data += cc.data
vispred = predict(vispred, model, **kwargs)
visres.data['vis'] = vis.data['vis'] - vispred.data['vis']
dirty, sumwt = invert(visres, model, **kwargs)
if numpy.abs(dirty.data).max() < 1.1 * thresh:
log.info("Reached stopping threshold %.6f Jy" % thresh)
break
log.info("solve_image: End of major cycle")
log.info("solve_image: End of major cycles")
return visres, model, dirty
def vis_timeslice_iter(vis, **kwargs):
""" Time slice iterator
If timeslice='auto' then timeslice is taken to be the difference between the first two
unique elements of the vis time.
:param timeslice: Timeslice (seconds) ('auto')
:returns: Boolean array with selected rows=True
"""
assert type(vis) == Visibility or type(vis) == BlockVisibility
uniquetimes = numpy.unique(vis.time)
timeslice = get_parameter(kwargs, "timeslice", 'auto')
if timeslice == 'auto':
log.debug('vis_timeslice_iter: Found %d unique times' %
len(uniquetimes))
if len(uniquetimes) > 1:
timeslice = (uniquetimes[1] - uniquetimes[0])
log.debug(
'vis_timeslice_auto: Guessing time interval to be %.2f s' %
timeslice)
else:
# Doesn't matter what we set it to.
timeslice = vis.integration_time[0]
boxes = timeslice * numpy.round(uniquetimes / timeslice).astype('int')
for box in boxes:
rows = numpy.abs(vis.time - box) < 0.5 * timeslice
yield rows
def qa_image(im, mask=None, **kwargs) -> QA:
"""Assess the quality of an image
:param params:
:param im:
:return: QA
"""
assert type(im) == Image
if mask is None:
data = {'shape': str(im.data.shape),
'max': numpy.max(im.data),
'min': numpy.min(im.data),
'rms': numpy.std(im.data),
'sum': numpy.sum(im.data),
'medianabs': numpy.median(numpy.abs(im.data)),
'median': numpy.median(im.data)}
else:
mdata = im.data[mask.data > 0.0]
data = {'shape': str(im.data.shape),
'max': numpy.max(mdata),
'min': numpy.min(mdata),
'rms': numpy.std(mdata),
'sum': numpy.sum(mdata),
'medianabs': numpy.median(numpy.abs(mdata)),
'median': numpy.median(mdata)}
qa = QA(origin="qa_image",
data=data,
context=get_parameter(kwargs, 'context', ""))
return qa
def invert_2d_from_grid(uv_grid_vis,
sumwt,
im: Image,
normalize: bool = True,
**kwargs):
"""
Perform the 2d fourier transform on the gridded uv visibility
:param uv_grid_vis: the gridded uv visibility
:param sumwt: the weight for the uv visibility
:param im: image template (not changed)
:param return: resulting image[nchan, npol, ny, nx], sum of weights[nchan, npol]
"""
### Calculate gcf -> grid correction function ###
# 2D Prolate spheroidal angular function is separable
npixel = get_parameter(kwargs, "npixel", 512)
nx = npixel
ny = npixel
nu = numpy.abs(2.0 * (numpy.arange(nx) - nx // 2) / nx)
gcf1d, _ = grdsf(nu)
gcf = numpy.outer(gcf1d, gcf1d)
gcf[gcf > 0.0] = gcf.max() / gcf[gcf > 0.0]
result = numpy.real(ifft(uv_grid_vis)) * gcf
#Create image array
resultimage = create_image_from_array(result, im.wcs)
if normalize:
resultimage = normalize_sumwt(resultimage, sumwt)
return resultimage, sumwt
def do_2d_grid_only(vis: Visibility,
im: Image,
dopsf=False,
normalize=True,
only_gridding: bool = True,
**kwargs):
""" Do the gridding using prolate spheroidal gridding function
Use the image im as a template. Do PSF in a separate call.
Note that the image is not normalised but the sum of the weights. This is for ease of use in partitioning.
:param vis: Visibility to be inverted
:param im: image template (not changed)
:param dopsf: Make the psf instead of the dirty image
:param normalize: Normalize by the sum of weights (True)
:param return: grid
"""
log.debug("do_2d_grid_only: create grid only")
kwargs['kernel'] = get_parameter(kwargs, "kernel", '2d')
return invert_2d_base(vis,
im,
dopsf,
normalize=normalize,
only_gridding=only_gridding,
**kwargs)
def calibrate_visibility(vt: Visibility,
model: Image = None,
components=None,
predict=predict_2d,
**kwargs) -> Visibility:
""" calibrate Visibility with respect to model and optionally components
:param vt: Visibility
:param model: Model image
:param components: Sky components
:return: Calibrated visibility
"""
assert model is not None or components is not None, "calibration requires a model or skycomponents"
vtpred = copy_visibility(vt, zero=True)
if model is not None:
vtpred = predict(vtpred, model, **kwargs)
if components is not None:
vtpred = predict_skycomponent_visibility(vtpred, components)
else:
vtpred = predict_skycomponent_visibility(vtpred, components)
bvt = decoalesce_visibility(vt)
bvtpred = decoalesce_visibility(vtpred)
gt = solve_gaintable(bvt, bvtpred, **kwargs)
bvt = apply_gaintable(bvt,
gt,
inverse=get_parameter(kwargs, "inverse", False))
return convert_blockvisibility_to_visibility(bvt)
def image_raster_iter(im: Image, **kwargs):
"""Create an image_raster_iter generator, returning images, optionally with overlaps
The WCS is adjusted appropriately for each raster element. Hence this is a coordinate-aware
way to iterate through an image.
Provided we don't break reference semantics, memory should be conserved
To update the image in place:
for r in raster(im, facets=2)::
r.data[...] = numpy.sqrt(r.data[...])
:param im: Image
:param facets: Number of image partitions on each axis (2)
:param overlap: overlap in pixels
:param kwargs: throw away unwanted parameters
"""
nchan, npol, ny, nx = im.shape
facets = get_parameter(kwargs, "facets", 1)
overlap = get_parameter(kwargs, 'overlap', 0)
log.debug("raster_overlap: predicting using %d x %d image partitions" %
(facets, facets))
assert facets <= ny, "Cannot have more raster elements than pixels"
assert facets <= nx, "Cannot have more raster elements than pixels"
sx = int((nx // facets))
sy = int((ny // facets))
dx = int((nx // facets) + 2 * overlap)
dy = int((ny // facets) + 2 * overlap)
log.debug('raster_overlap: spacing of raster (%d, %d)' % (dx, dy))
for fy in range(facets):
y = ny // 2 + sy * (fy - facets // 2) - overlap
for fx in range(facets):
x = nx // 2 + sx * (fx - facets // 2) - overlap
if (x >= 0) and (x + dx) <= nx and (y >= 0) and (y + dy) <= ny:
log.debug('raster_overlap: partition (%d, %d) of (%d, %d)' %
(fy, fx, facets, facets))
# Adjust WCS
wcs = im.wcs.deepcopy()
wcs.wcs.crpix[0] -= x
wcs.wcs.crpix[1] -= y
# yield image from slice (reference!)
yield create_image_from_array(im.data[..., y:y + dy, x:x + dx],
wcs, im.polarisation_frame)
def predict_2d_base(vis: Union[BlockVisibility, Visibility], model: Image,
**kwargs) -> Union[BlockVisibility, Visibility]:
""" Predict using convolutional degridding.
This is at the bottom of the layering i.e. all transforms are eventually expressed in terms of
this function. Any shifting needed is performed here.
:param vis: Visibility to be predicted
:param model: model image
:return: resulting visibility (in place works)
"""
if isinstance(vis, BlockVisibility):
log.debug("imaging.predict: coalescing prior to prediction")
avis = coalesce_visibility(vis, **kwargs)
else:
avis = vis
assert isinstance(avis, Visibility), avis
_, _, ny, nx = model.data.shape
padding = {}
if get_parameter(kwargs, "padding", False):
padding = {'padding': get_parameter(kwargs, "padding", False)}
spectral_mode, vfrequencymap = get_frequency_map(avis, model)
polarisation_mode, vpolarisationmap = get_polarisation_map(avis, model)
uvw_mode, shape, padding, vuvwmap = get_uvw_map(avis, model, **padding)
kernel_name, gcf, vkernellist = get_kernel_list(avis, model, **kwargs)
uvgrid = fft((pad_mid(model.data, int(round(padding * nx))) *
gcf).astype(dtype=complex))
avis.data['vis'] = convolutional_degrid(vkernellist,
avis.data['vis'].shape, uvgrid,
vuvwmap, vfrequencymap,
vpolarisationmap)
# Now we can shift the visibility from the image frame to the original visibility frame
svis = shift_vis_to_image(avis, model, tangent=True, inverse=True)
if isinstance(vis, BlockVisibility) and isinstance(svis, Visibility):
log.debug("imaging.predict decoalescing post prediction")
return decoalesce_visibility(svis)
else:
return svis
def vis_slice_iter(vis, **kwargs):
""" Iterates in slices
:param step: Size of step to be iterated over (in rows)
:param vis_slices: Number of slices (second in precedence to step)
:returns: Boolean array with selected rows=True
"""
assert type(vis) == Visibility or type(vis) == BlockVisibility
step = get_parameter(kwargs, "step", None)
if step is None:
vis_slices = get_parameter(kwargs, "vis_slices", 1)
step = 1 + vis.nvis // vis_slices
assert step > 0
for row in range(0, vis.nvis, step):
yield range(row, min(row + step, vis.nvis))
def predict_2d(vis: Visibility, im: Image, **kwargs) -> Visibility:
""" Predict using convolutional degridding and w projection
:param vis: Visibility to be predicted
:param model: model image
:return: resulting visibility (in place works)
"""
log.debug("predict_2d: predict using 2d transform")
timing = get_parameter(kwargs, "timing", False)
if timing:
return predict_2d_base_timing(vis, im, **kwargs)
return predict_2d_base(vis, im, **kwargs)
def t1(**kwargs):
assert get_parameter(kwargs, 'cellsize') == 0.1
assert get_parameter(kwargs, 'spectral_mode', 'channels') == 'mfs'
assert get_parameter(kwargs, 'null_mode', 'mfs') == 'mfs'
assert get_parameter(kwargs, 'foo', 'bar') == 'bar'
assert get_parameter(kwargs, 'foo') is None
assert get_parameter(None, 'foo', 'bar') == 'bar'
def create_deconvolve_graph(dirty_graph: delayed, psf_graph: delayed,
model_graph: delayed, **kwargs) -> delayed:
"""Create a graph for deconvolution, adding to the model
:param dirty_graph:
:param psf_graph:
:param model_graph:
:param nchan: Number of channels
:param kwargs: Parameters for functions in graphs
:return:
"""
nchan = get_parameter(kwargs, "nchan", 1)
def remove_sumwt(dirty_list):
return [d[0] for d in dirty_list]
def make_cube_and_deconvolve(dirty, psf, model):
dirty_cube = image_gather_channels(remove_sumwt(dirty),
subimages=nchan)
psf_cube = image_gather_channels(remove_sumwt(psf), subimages=nchan)
model_cube = image_gather_channels(model, subimages=nchan)
result = deconvolve_cube(dirty_cube, psf_cube, **kwargs)
result[0].data += model_cube.data
return image_scatter_channels(result[0], nchan)
def deconvolve(dirty, psf, model):
result = deconvolve_cube(dirty[0], psf[0], **kwargs)
result[0].data += model.data
return result[0]
algorithm = get_parameter(kwargs, "algorithm", 'mmclean')
if algorithm == "mmclean" and nchan > 1:
return delayed(make_cube_and_deconvolve,
nout=nchan)(dirty_graph, psf_graph, model_graph)
else:
return [
delayed(deconvolve, pure=True,
nout=nchan)(dirty_graph[i], psf_graph[i], model_graph[i])
for i, _ in enumerate(dirty_graph)
]
def vis_slice_iter(vis: Union[Visibility, BlockVisibility],
**kwargs) -> numpy.ndarray:
""" Iterates in slices
:param step: Size of step to be iterated over (in rows)
:param vis_slices: Number of slices (second in precedence to step)
:return: Boolean array with selected rows=True
"""
assert isinstance(vis, Visibility) or isinstance(vis, BlockVisibility)
step = get_parameter(kwargs, "step", None)
if step is None:
vis_slices = get_parameter(kwargs, "vis_slices", 1)
if isinstance(vis_slices, int):
step = vis.nvis // vis_slices
else:
step = vis.nvis
assert step > 0
for row in range(0, vis.nvis, step):
yield range(row, min(row + step, vis.nvis))
def vis_slice_iter(vis: Union[Visibility, BlockVisibility],
**kwargs) -> numpy.ndarray:
""" Iterates in slices
:param step: Size of step to be iterated over (in rows)
:param vis_slices: Number of slices (second in precedence to step)
:return: Boolean array with selected rows=True
"""
assert isinstance(vis, Visibility) or isinstance(vis, BlockVisibility), vis
step = get_parameter(kwargs, "step", None)
if step is None:
vis_slices = get_parameter(kwargs, "vis_slices", None)
assert vis_slices is not None, "vis slicing not specified: set either step or vis_slices"
step = vis.nvis // vis_slices
assert step > 0
for row in range(0, vis.nvis, step):
rows = vis.nvis * [False]
for r in range(row, min(row + step, vis.nvis)):
rows[r] = True
yield rows
def vis_timeslice_iter(vis: Visibility, **kwargs) -> numpy.ndarray:
""" W slice iterator
:param timeslice: timeslice (arcseconds ???)
:param vis_slices: Number of slices
:return: Boolean array with selected rows=True
"""
assert isinstance(vis, Visibility) or isinstance(vis, BlockVisibility)
timemin = numpy.min(vis.time)
timemax = numpy.max(vis.time)
timeslice = get_parameter(kwargs, "timeslice", None)
if timeslice is None or timeslice == 'auto':
vis_slices = get_parameter(kwargs, "vis_slices", None)
if vis_slices is None or vis_slices == 'auto':
vis_slices = len(numpy.unique(vis.time))
### Change code to equaly slice the data ###
#boxes = numpy.linspace(timemin, timemax, vis_slices)
half_box_width = (timemax - timemin) / (2 * vis_slices)
boxes = numpy.linspace(timemin + half_box_width,
timemax - half_box_width, vis_slices)
#Equal slicing as now linspace distribute points to the middle of boxes
timeslice = (timemax - timemin) / vis_slices
else:
vis_slices = 1 + 2 * numpy.ceil(
(timemax - timemin) / timeslice).astype('int')
boxes = numpy.linspace(timemin, timemax, vis_slices)
if vis_slices > 1:
timeslice = boxes[1] - boxes[0]
else:
timeslice = timemax - timemin
for box in boxes:
rows = numpy.abs(vis.time - box) <= 0.5 * timeslice
yield rows
def get_kernel_list(vis: Visibility, im: Image, **kwargs):
"""Get the list of kernels, one per visibility
"""
shape = im.data.shape
npixel = shape[3]
cellsize = numpy.pi * im.wcs.wcs.cdelt[1] / 180.0
kernelname = get_parameter(kwargs, "kernel", "2d")
oversampling = get_parameter(kwargs, "oversampling", 8)
padding = get_parameter(kwargs, "padding", 2)
gcf, _ = anti_aliasing_calculate((padding * npixel, padding * npixel), oversampling)
wabsmax = numpy.max(numpy.abs(vis.w))
if kernelname == 'wprojection' and wabsmax > 0.0:
# wprojection needs a lot of commentary!
log.debug("get_kernel_list: Using wprojection kernel")
# The field of view must be as padded! R_F is for reporting only so that
# need not be padded.
fov = cellsize * npixel * padding
r_f = (cellsize * npixel / 2) ** 2 / abs(cellsize)
log.debug("get_kernel_list: Fresnel number = %f" % (r_f))
delA = get_parameter(kwargs, 'wloss', 0.02)
advice = advise_wide_field(vis, delA)
wstep = get_parameter(kwargs, "wstep", advice['w_sampling_primary_beam'])
log.debug("get_kernel_list: Using w projection with wstep = %f" % (wstep))
# Now calculate the maximum support for the w kernel
kernelwidth = get_parameter(kwargs, "kernelwidth",
(2 * int(round(numpy.sin(0.5 * fov) * npixel * wabsmax * cellsize))))
kernelwidth = max(kernelwidth, 8)
assert kernelwidth % 2 == 0
log.debug("get_kernel_list: Maximum w kernel full width = %d pixels" % (kernelwidth))
padded_shape=[im.shape[0], im.shape[1], im.shape[2] * padding, im.shape[3] * padding]
remove_shift = get_parameter(kwargs, "remove_shift", True)
padded_image = pad_image(im, padded_shape)
kernel_list = w_kernel_list(vis, padded_image, oversampling=oversampling, wstep=wstep,
kernelwidth=kernelwidth, remove_shift=remove_shift)
else:
kernelname = '2d'
kernel_list = standard_kernel_list(vis, (padding * npixel, padding * npixel),
oversampling=oversampling)
return kernelname, gcf, kernel_list
def restore_cube(model: Image, psf: Image, residual=None, **kwargs) -> Image:
""" Restore the model image to the residuals
:params psf: Input PSF
:return: restored image
"""
assert isinstance(model, Image), "Type is %s" % (type(model))
assert isinstance(psf, Image), "Type is %s" % (type(psf))
assert residual is None or isinstance(
residual, Image), "Type is %s" % (type(residual))
restored = copy_image(model)
npixel = psf.data.shape[3]
sl = slice(npixel // 2 - 7, npixel // 2 + 8)
size = get_parameter(kwargs, "psfwidth", None)
if size is None:
# isotropic at the moment!
try:
fit = fit_2dgaussian(psf.data[0, 0, sl, sl])
if fit.x_stddev <= 0.0 or fit.y_stddev <= 0.0:
log.debug(
'restore_cube: error in fitting to psf, using 1 pixel stddev'
)
size = 1.0
else:
size = max(fit.x_stddev, fit.y_stddev)
log.debug('restore_cube: psfwidth = %s' % (size))
except:
log.debug(
'restore_cube: warning in fit to psf, using 1 pixel stddev')
size = 1.0
else:
log.debug('restore_cube: Using specified psfwidth = %s' % (size))
# By convention, we normalise the peak not the integral so this is the volume of the Gaussian
norm = 2.0 * numpy.pi * size**2
gk = Gaussian2DKernel(size)
for chan in range(model.shape[0]):
for pol in range(model.shape[1]):
restored.data[chan, pol, :, :] = norm * convolve(
model.data[chan, pol, :, :], gk, normalize_kernel=False)
if residual is not None:
restored.data += residual.data
return restored
def create_ical_pipeline_graph(vis_graph_list,
model_graph,
c_deconvolve_graph=create_deconvolve_graph,
c_invert_graph=create_invert_graph,
c_residual_graph=create_residual_graph,
first_selfcal=None,
**kwargs):
"""Create graph for ICAL pipeline
:param vis_graph_list:
:param model_graph:
:param c_deconvolve_graph: Default: create_deconvolve_graph
:param c_invert_graph: Default: create_invert_graph,
:param c_residual_graph: Default: Default: create_residual graph
:param kwargs:
:return:
"""
psf_graph = c_invert_graph(vis_graph_list,
model_graph,
dopsf=True,
**kwargs)
if first_selfcal is not None and first_selfcal == 0:
vis_graph_list = create_selfcal_graph_list(vis_graph_list, model_graph,
**kwargs)
residual_graph = c_residual_graph(vis_graph_list, model_graph, **kwargs)
deconvolve_model_graph = c_deconvolve_graph(residual_graph, psf_graph,
model_graph, **kwargs)
nmajor = get_parameter(kwargs, "nmajor", 5)
if nmajor > 1:
for cycle in range(nmajor):
if first_selfcal is not None and cycle >= first_selfcal:
vis_graph_list = create_selfcal_graph_list(
vis_graph_list, deconvolve_model_graph, **kwargs)
residual_graph = c_residual_graph(vis_graph_list,
deconvolve_model_graph, **kwargs)
deconvolve_model_graph = c_deconvolve_graph(
residual_graph, psf_graph, deconvolve_model_graph, **kwargs)
residual_graph = c_residual_graph(vis_graph_list,
deconvolve_model_graph, **kwargs)
restore_graph = delayed(restore_cube, pure=True,
nout=1)(deconvolve_model_graph, psf_graph[0],
residual_graph[0], **kwargs)
return delayed((deconvolve_model_graph, residual_graph, restore_graph))
def create_continuum_imaging_pipeline_graph(vis_graph_list,
model_graph: delayed,
context='2d',
**kwargs) -> delayed:
""" Create graph for the continuum imaging pipeline.
Same as ICAL but with no selfcal.
:param vis_graph_list:
:param model_graph:
:param c_deconvolve_graph: Default: create_deconvolve_graph
:param c_invert_graph: Default: create_invert_graph
:param c_residual_graph: Default: Default: create_residual graph
:param kwargs: Parameters for functions in graphs
:return:
"""
psf_graph = create_invert_graph(vis_graph_list,
model_graph,
dopsf=True,
context=context,
**kwargs)
residual_graph = create_residual_graph(vis_graph_list,
model_graph,
context=context,
**kwargs)
deconvolve_model_graph = create_deconvolve_graph(residual_graph, psf_graph,
model_graph, **kwargs)
nmajor = get_parameter(kwargs, "nmajor", 5)
if nmajor > 1:
for cycle in range(nmajor):
residual_graph = create_residual_graph(vis_graph_list,
deconvolve_model_graph,
context=context,
**kwargs)
deconvolve_model_graph = create_deconvolve_graph(
residual_graph, psf_graph, deconvolve_model_graph, **kwargs)
residual_graph = create_residual_graph(vis_graph_list,
deconvolve_model_graph,
context=context,
**kwargs)
restore_graph = create_restore_graph(deconvolve_model_graph, psf_graph,
residual_graph)
return delayed((deconvolve_model_graph, residual_graph, restore_graph))
def get_uvw_map(vis, im, **kwargs):
""" Get the generators that map channels uvw to pixels
"""
# Transform parameters
padding = get_parameter(kwargs, "padding", 2)
# Model image information
inchan, inpol, ny, nx = im.data.shape
shape = (1, int(round(padding * ny)), int(round(padding * nx)))
# UV sampling information
uvwscale = numpy.zeros([3])
uvwscale[0:2] = im.wcs.wcs.cdelt[0:2] * numpy.pi / 180.0
assert uvwscale[0] != 0.0, "Error in uv scaling"
vuvwmap = uvwscale * vis.uvw
uvw_mode = "2d"
return uvw_mode, shape, padding, vuvwmap
