def create_low_test_beam(model: Image) -> Image:
"""Create a test power beam for LOW using an image from OSKAR
:param model: Template image
:return: Image
"""
beam = import_image_from_fits(arl_path('data/models/SKA1_LOW_beam.fits'))
# Scale the image cellsize to account for the different in frequencies. Eventually we will want to
# use a frequency cube
log.info("create_low_test_beam: primary beam is defined at %.3f MHz" %
(beam.wcs.wcs.crval[2] * 1e-6))
nchan, npol, ny, nx = model.shape
# We need to interpolate each frequency channel separately. The beam is assumed to just scale with
# frequency.
reprojected_beam = create_empty_image_like(model)
for chan in range(nchan):
model2dwcs = model.wcs.sub(2).deepcopy()
model2dshape = [model.shape[2], model.shape[3]]
beam2dwcs = beam.wcs.sub(2).deepcopy()
# The frequency axis is the second to last in the beam
frequency = model.wcs.sub(['spectral']).wcs_pix2world([chan], 0)[0]
fscale = beam.wcs.wcs.crval[2] / frequency
beam2dwcs.wcs.cdelt = fscale * beam.wcs.sub(2).wcs.cdelt
beam2dwcs.wcs.crpix = beam.wcs.sub(2).wcs.crpix
beam2dwcs.wcs.crval = model.wcs.sub(2).wcs.crval
beam2dwcs.wcs.ctype = model.wcs.sub(2).wcs.ctype
model2dwcs.wcs.crpix = [
model.shape[2] // 2 + 1, model.shape[3] // 2 + 1
]
beam2d = create_image_from_array(beam.data[0, 0, :, :], beam2dwcs,
model.polarisation_frame)
reprojected_beam2d, footprint = reproject_image(beam2d,
model2dwcs,
shape=model2dshape)
assert numpy.max(
footprint.data) > 0.0, "No overlap between beam and model"
reprojected_beam2d.data *= reprojected_beam2d.data
reprojected_beam2d.data[footprint.data <= 0.0] = 0.0
for pol in range(npol):
reprojected_beam.data[chan,
pol, :, :] = reprojected_beam2d.data[:, :]
return reprojected_beam
def test_reproject(self):
# Reproject an image
cellsize = 1.5 * self.cellsize
newwcs = self.m31image.wcs.deepcopy()
newwcs.wcs.cdelt[0] = -cellsize
newwcs.wcs.cdelt[1] = +cellsize
newshape = numpy.array(self.m31image.data.shape)
newshape[2] /= 1.5
newshape[3] /= 1.5
newimage, footprint = reproject_image(self.m31image,
newwcs,
shape=newshape)
checkwcs(newimage.wcs, newwcs)
def predict_timeslice_single(vis: Visibility,
model: Image,
predict=predict_2d,
remove=True,
gcfcf=None,
**kwargs) -> Visibility:
""" Predict using a single time slices.
This fits a single plane and corrects the image geometry.
:param vis: Visibility to be predicted
:param model: model image
:param predict:
:param remove: Remove fitted w (so that wprojection will do the right thing)
:param gcfcf: (Grid correction function, convolution function)
:return: resulting visibility (in place works)
"""
assert isinstance(vis, Visibility), vis
vis.data['vis'][...] = 0.0
# Fit and remove best fitting plane for this slice
uvw = vis.uvw
avis, p, q = fit_uvwplane(vis, remove=remove)
# We want to describe work image as distorted. We describe the distortion by putting
# the olbiquity parameters in the wcs. The input model should be described as having
# zero olbiquity parameters.
# Note that this has to be zero relative in first element, one relative in second!!!
if numpy.abs(p) > 1e-7 or numpy.abs(q) > 1e-7:
newwcs = model.wcs.deepcopy()
newwcs.wcs.set_pv([(0, 1, -p), (0, 2, -q)])
workimage, footprintimage = reproject_image(model,
newwcs,
shape=model.shape)
workimage.data[footprintimage.data <= 0.0] = 0.0
workimage.wcs.wcs.set_pv([(0, 1, -p), (0, 2, -q)])
# Now we can do the predict
vis = predict(avis, workimage, gcfcf=gcfcf, **kwargs)
else:
vis = predict(avis, model, gcfcf=gcfcf, **kwargs)
if remove:
avis.data['uvw'][...] = uvw
return vis
def invert_timeslice_single(vis: Visibility,
im: Image,
dopsf,
normalize=True,
remove=True,
gcfcf=None,
**kwargs) -> (Image, numpy.ndarray):
"""Process single time slice
Extracted for re-use in parallel version
:param vis: Visibility to be inverted
:param im: image template (not changed)
:param dopsf: Make the psf instead of the dirty image
:param gcfcf: (Grid correction function, convolution function)
:param normalize: Normalize by the sum of weights (True)
"""
assert isinstance(vis, Visibility), vis
uvw = vis.uvw
vis, p, q = fit_uvwplane(vis, remove=remove)
workimage, sumwt = invert_2d(vis,
im,
dopsf,
normalize=normalize,
gcfcf=gcfcf,
**kwargs)
# Work image is distorted. We describe the distortion by putting the olbiquity parameters in
# the wcs. The output image should be described as having zero olbiquity parameters.
if numpy.abs(p) > 1e-7 or numpy.abs(q) > 1e-7:
# Note that this has to be zero relative in first element, one relative in second!!!!
workimage.wcs.wcs.set_pv([(0, 1, -p), (0, 2, -q)])
finalimage, footprint = reproject_image(workimage, im.wcs, im.shape)
finalimage.data[footprint.data <= 0.0] = 0.0
finalimage.wcs.wcs.set_pv([(0, 1, 0.0), (0, 2, 0.0)])
if remove:
vis.data['uvw'][...] = uvw
return finalimage, sumwt
else:
if remove:
vis.data['uvw'][...] = uvw
return workimage, sumwt
def create_awterm_convolutionfunction(im,
make_pb=None,
nw=1,
wstep=1e15,
oversampling=8,
support=6,
use_aaf=True,
maxsupport=512):
""" Fill AW projection kernel into a GridData.
:param im: Image template
:param make_pb: Function to make the primary beam model image
:param nw: Number of w planes
:param wstep: Step in w (wavelengths)
:param oversampling: Oversampling of the convolution function in uv space
:return: griddata correction Image, griddata kernel as GridData
"""
d2r = numpy.pi / 180.0
# We only need the griddata correction function for the PSWF so we make
# it for the shape of the image
nchan, npol, ony, onx = im.data.shape
assert isinstance(im, Image)
# Calculate the template convolution kernel.
cf = create_convolutionfunction_from_image(im,
oversampling=oversampling,
support=support)
cf_shape = list(cf.data.shape)
cf_shape[2] = nw
cf.data = numpy.zeros(cf_shape).astype('complex')
cf.grid_wcs.wcs.crpix[4] = nw // 2 + 1.0
cf.grid_wcs.wcs.cdelt[4] = wstep
cf.grid_wcs.wcs.ctype[4] = 'WW'
if numpy.abs(wstep) > 0.0:
w_list = cf.grid_wcs.sub([5]).wcs_pix2world(range(nw), 0)[0]
else:
w_list = [0.0]
assert isinstance(oversampling, int)
assert oversampling > 0
nx = max(maxsupport, 2 * oversampling * support)
ny = max(maxsupport, 2 * oversampling * support)
qnx = nx // oversampling
qny = ny // oversampling
cf.data[...] = 0.0
subim = copy_image(im)
ccell = onx * numpy.abs(d2r * subim.wcs.wcs.cdelt[0]) / qnx
subim.data = numpy.zeros([nchan, npol, qny, qnx])
subim.wcs.wcs.cdelt[0] = -ccell / d2r
subim.wcs.wcs.cdelt[1] = +ccell / d2r
subim.wcs.wcs.crpix[0] = qnx // 2 + 1.0
subim.wcs.wcs.crpix[1] = qny // 2 + 1.0
if use_aaf:
this_pswf_gcf, _ = create_pswf_convolutionfunction(subim,
oversampling=1,
support=6)
norm = 1.0 / this_pswf_gcf.data
else:
norm = 1.0
if make_pb is not None:
pb = make_pb(subim)
rpb, footprint = reproject_image(pb, subim.wcs, shape=subim.shape)
rpb.data[footprint.data < 1e-6] = 0.0
norm *= rpb.data
# We might need to work with a larger image
padded_shape = [nchan, npol, ny, nx]
thisplane = copy_image(subim)
thisplane.data = numpy.zeros(thisplane.shape, dtype='complex')
for z, w in enumerate(w_list):
thisplane.data[...] = 0.0 + 0.0j
thisplane = create_w_term_like(thisplane, w, dopol=True)
thisplane.data *= norm
paddedplane = pad_image(thisplane, padded_shape)
paddedplane = fft_image(paddedplane)
ycen, xcen = ny // 2, nx // 2
for y in range(oversampling):
ybeg = y + ycen + (support * oversampling) // 2 - oversampling // 2
yend = y + ycen - (support * oversampling) // 2 - oversampling // 2
vv = range(ybeg, yend, -oversampling)
for x in range(oversampling):
xbeg = x + xcen + (support *
oversampling) // 2 - oversampling // 2
xend = x + xcen - (support *
oversampling) // 2 - oversampling // 2
uu = range(xbeg, xend, -oversampling)
for chan in range(nchan):
for pol in range(npol):
cf.data[chan, pol, z, y, x, :, :] = paddedplane.data[
chan, pol, :, :][vv, :][:, uu]
cf.data /= numpy.sum(
numpy.real(cf.data[0, 0, nw // 2, oversampling // 2,
oversampling // 2, :, :]))
cf.data = numpy.conjugate(cf.data)
if use_aaf:
pswf_gcf, _ = create_pswf_convolutionfunction(im,
oversampling=1,
support=6)
else:
pswf_gcf = create_empty_image_like(im)
pswf_gcf.data[...] = 1.0
return pswf_gcf, cf
def create_awterm_convolutionfunction(im,
make_pb=None,
nw=1,
wstep=1e15,
oversampling=8,
support=6,
use_aaf=True,
maxsupport=512,
**kwargs):
""" Fill AW projection kernel into a GridData.
:param im: Image template
:param make_pb: Function to make the primary beam model image (hint: use a partial)
:param nw: Number of w planes
:param wstep: Step in w (wavelengths)
:param oversampling: Oversampling of the convolution function in uv space
:return: griddata correction Image, griddata kernel as GridData
"""
d2r = numpy.pi / 180.0
# We only need the griddata correction function for the PSWF so we make
# it for the shape of the image
nchan, npol, ony, onx = im.data.shape
assert isinstance(im, Image)
# Calculate the template convolution kernel.
cf = create_convolutionfunction_from_image(im,
oversampling=oversampling,
support=support)
cf_shape = list(cf.data.shape)
cf_shape[2] = nw
cf.data = numpy.zeros(cf_shape).astype('complex')
cf.grid_wcs.wcs.crpix[4] = nw // 2 + 1.0
cf.grid_wcs.wcs.cdelt[4] = wstep
cf.grid_wcs.wcs.ctype[4] = 'WW'
if numpy.abs(wstep) > 0.0:
w_list = cf.grid_wcs.sub([5]).wcs_pix2world(range(nw), 0)[0]
else:
w_list = [0.0]
assert isinstance(oversampling, int)
assert oversampling > 0
nx = max(maxsupport, 2 * oversampling * support)
ny = max(maxsupport, 2 * oversampling * support)
qnx = nx // oversampling
qny = ny // oversampling
cf.data[...] = 0.0
subim = copy_image(im)
ccell = onx * numpy.abs(d2r * subim.wcs.wcs.cdelt[0]) / qnx
subim.data = numpy.zeros([nchan, npol, qny, qnx])
subim.wcs.wcs.cdelt[0] = -ccell / d2r
subim.wcs.wcs.cdelt[1] = +ccell / d2r
subim.wcs.wcs.crpix[0] = qnx // 2 + 1.0
subim.wcs.wcs.crpix[1] = qny // 2 + 1.0
if use_aaf:
this_pswf_gcf, _ = create_pswf_convolutionfunction(subim,
oversampling=1,
support=6)
norm = 1.0 / this_pswf_gcf.data
else:
norm = 1.0
if make_pb is not None:
pb = make_pb(subim)
rpb, footprint = reproject_image(pb, subim.wcs, shape=subim.shape)
rpb.data[footprint.data < 1e-6] = 0.0
norm *= rpb.data
# We might need to work with a larger image
padded_shape = [nchan, npol, ny, nx]
thisplane = copy_image(subim)
thisplane.data = numpy.zeros(thisplane.shape, dtype='complex')
for z, w in enumerate(w_list):
thisplane.data[...] = 0.0 + 0.0j
thisplane = create_w_term_like(thisplane, w, dopol=True)
thisplane.data *= norm
paddedplane = pad_image(thisplane, padded_shape)
paddedplane = fft_image(paddedplane)
ycen, xcen = ny // 2, nx // 2
for y in range(oversampling):
ybeg = y + ycen + (support * oversampling) // 2 - oversampling // 2
yend = y + ycen - (support * oversampling) // 2 - oversampling // 2
# vv = range(ybeg, yend, -oversampling)
for x in range(oversampling):
xbeg = x + xcen + (support *
oversampling) // 2 - oversampling // 2
xend = x + xcen - (support *
oversampling) // 2 - oversampling // 2
# uu = range(xbeg, xend, -oversampling)
cf.data[..., z, y,
x, :, :] = paddedplane.data[...,
ybeg:yend:-oversampling,
xbeg:xend:-oversampling]
# for chan in range(nchan):
# for pol in range(npol):
# cf.data[chan, pol, z, y, x, :, :] = paddedplane.data[chan, pol, :, :][vv, :][:, uu]
cf.data /= numpy.sum(
numpy.real(cf.data[0, 0, nw // 2, oversampling // 2,
oversampling // 2, :, :]))
cf.data = numpy.conjugate(cf.data)
#====================================
#Use ASKAPSoft routine to crop the support size
crop_ASKAPSOft_like = True
if crop_ASKAPSOft_like:
#Hardcode the cellsize: 1 / FOV
#uv_cellsize = 57.3;#N=1200 pixel and pixelsize is 3 arcseconds
#uv_cellsize = 43.97;#N=1600 pixel and pixelsize is 3 arcseconds
#uv_cellsize = 114.6;#N=1800 pixel with 1 arcsecond pixelsize
#uv_cellsize = 57.3;#N=1800 pixel with 2 arcsecond pixelsize
#uv_cellsize = 1145.91509915;#N=1800 pixxel with 0.1 arcsecond pixelsize
#Get from **kwargs
if kwargs is None:
#Safe solution works for baselines up to > 100km and result in small kernels
uv_cellsize = 1145.91509915
#N=1800 pixxel with 0.1 arcsecond pixelsize
if 'UVcellsize' in kwargs.keys():
uv_cellsize = kwargs['UVcellsize']
#print(uv_cellsize);
#Cutoff param in ASKAPSoft hardcoded as well
ASKAPSoft_cutof = 0.1
wTheta_list = numpy.zeros(len(w_list))
for i in range(0, len(w_list)):
if w_list[i] == 0:
wTheta_list[i] = 0.9
#This is due to the future if statements cause if it is small, the kernel will be 3 which is a clear cutoff
else:
wTheta_list[i] = numpy.fabs(
w_list[i]) / (uv_cellsize * uv_cellsize)
kernel_size_list = []
#We rounded the kernels according to conventional rounding rules
for i in range(0, len(wTheta_list)):
#if wTheta_list[i] < 1:
if wTheta_list[i] < 1: #Change to ASKAPSoft
kernel_size_list.append(int(3.))
elif ASKAPSoft_cutof < 0.01:
kernel_size_list.append(int(6 + 1.14 * wTheta_list[i]))
else:
kernel_size_list.append(
int(numpy.sqrt(49 + wTheta_list[i] * wTheta_list[i])))
log.info('W-kernel w-terms:')
log.info(w_list)
log.info('Corresponding w-kernel sizes:')
log.info(kernel_size_list)
print(numpy.unique(kernel_size_list))
#print(kernel_size_list);
crop_list = []
#another rounding according to conventional rounding rules
for i in range(0, len(kernel_size_list)):
if support - kernel_size_list[i] <= 0:
crop_list.append(int(0))
else:
crop_list.append(int((support - kernel_size_list[i]) / 2))
#Crop original suppor
for i in range(0, nw):
if crop_list[i] != 0:
cf.data[0, 0, i, :, :, 0:crop_list[i], :] = 0
cf.data[0, 0, i, :, :, -crop_list[i]:, :] = 0
cf.data[0, 0, i, :, :, :, 0:crop_list[i]] = 0
cf.data[0, 0, i, :, :, :, -crop_list[i]:] = 0
else:
pass
#Plot
#import matplotlib.pyplot as plt
#cf.data[0,0,i,0,0,...][cf.data[0,0,i,0,0,...] != 0.] = 1+0.j;
#plt.imshow(numpy.real(cf.data[0,0,i,0,0,...]))
#plt.show(block=True)
#plt.close();
#====================================
if use_aaf:
pswf_gcf, _ = create_pswf_convolutionfunction(im,
oversampling=1,
support=6)
else:
pswf_gcf = create_empty_image_like(im)
pswf_gcf.data[...] = 1.0
return pswf_gcf, cf