def find_peaks(inimg="./ngvla_30dor/ngvla_30dor.ngvla-core-revC.skymodel.flat",
search_size=55,
thresh=4.5):
"""
Find peaks in a CASA image. respects image mask. Heuristic: do a local max filter over a
neighborhood of size search_size, and select those which are > 4.5 sigma (with sigma estimated
as MAD of the image as a whole)
"""
print "THRESH ", thresh
ia = iatool()
ia.open(inimg)
imvals = np.squeeze(ia.getchunk())
# somewhat confusingly, the mask is True where pixels are good, and False where bad.
immask = np.squeeze(ia.getchunk(getmask=True))
ia.close()
print imvals.shape
# compute some stats excluding NaN's -
#imvals2 = imvals[~np.isnan(imvals)]
imvals2 = imvals[np.isfinite(imvals) & immask]
pix_mean = imvals2.mean()
#pix_sd = np.std(imvals2)
print "MEDIAN"
pix_median = np.median(imvals2)
# note: by default this is normalized so that the std dev of
# a gaussian is 1.0
pix_mad = au.MAD(imvals2)
neighborhood = np.ones(
[np.round(search_size), np.round(search_size)], dtype=np.bool)
print neighborhood.shape
print "MAX Filter"
is_local_max = maximum_filter(imvals, footprint=neighborhood) == imvals
# this is a boolean array, same dims as imvals. True where you have a local max
# that's over the threshold specified here -
#highest_peaks = is_local_max & np.asarray(np.asarray(imvals > pix_median + 5.0*pix_mad))
# so is this -
print is_local_max.shape
highest_peaks = is_local_max & (imvals > (pix_median + thresh * pix_mad))
print highest_peaks.shape, pix_median, thresh, pix_mad
# this gives the indices where the value is TRUE -
peak_indices = np.argwhere(highest_peaks)
print peak_indices.shape
return peak_indices
def compare_fluxes(
img1="./ngvla_30dor/ngvla_30dor.ngvla-core-revC.autoDev2.image",
img2="./ngvla_30dor/ngvla_30dor.ngvla-core-revC.skymodel.flat",
mask1="",
mask2="",
search_size=2.0,
thresh=4.5,
search_img=1,
maxrad=50.0,
minrad=1.0,
radstep=1.0,
doAnnuli=False,
flipMask=False,
dumbTest=False):
"""
img1,img2: casa images to compare
search_size: width of search region in beams (beam is takem from chosen search image)
thresh: threshold for peak detection (robust sigma above median, default = 4.5)
search_img: which image (1 or 2) to search for peaks (must have a beam)
minrad,maxrad,radstep: start, stop, and delta radii for flux sums (in beams, again taken from search image)
images are presumed to be on the same grid to start with (shape, pixel size) -- use imgregrid task
and presumed to be in Jy/pixel (or otherwise same brightness unit per pixel) -- use combut.jyBm2jyPix()
returns: flux(search), flux(other), peak_locations
22jan2020 currently uses a square aperture (needs updating to circular) and hard wires the range of radii. return radii are in
pixels not arcsec.
"""
#print "THRESH ", thresh
if (search_img == 1):
main_img = img1
aux_img = img2
else:
main_img = img2
aux_img = img1
print " main image = ", main_img
ia = iatool()
# main image characteristics
ia.open(main_img)
# average over the stokes axis to get it down to 3 axes which is what our other one has
main_img_cs = ia.coordsys()
# how to trim the freq axis--
#img_shape = (ia.shape())[0:3]
main_img_shape = ia.shape()
print main_img_shape
imvals = np.squeeze(ia.getchunk())
immask = np.squeeze(ia.getchunk(getmask=True))
ia.close()
# get beam info
hdr = imhead(imagename=main_img, mode='summary')
cd1 = imhead(imagename=main_img, mode='get', hdkey='cdelt1')
cd1 = convert_angle(cd1, 'arcsec')
cd2 = imhead(imagename=main_img, mode='get', hdkey='cdelt2')
cd2 = convert_angle(cd2, 'arcsec')
mean_cell = (np.abs(cd1['value'] * cd2['value']))**0.5
bmaj_str = str(hdr['restoringbeam']['major']
['value']) + hdr['restoringbeam']['major']['unit']
bmin_str = str(hdr['restoringbeam']['minor']
['value']) + hdr['restoringbeam']['minor']['unit']
bpa_str = str(hdr['restoringbeam']['positionangle']
['value']) + hdr['restoringbeam']['positionangle']['unit']
bmaj = convert_angle(hdr['restoringbeam']['major'], 'arcsec')
bmin = convert_angle(hdr['restoringbeam']['minor'], 'arcsec')
beam_fwhm = (bmaj['value'] * bmin['value'])**0.5
print "beam: " + str(beam_fwhm) + " " + bmaj['unit']
print "mean pix: " + str(mean_cell) + " " + cd1['unit']
# secondary image characteristics
ia.open(aux_img)
# average over the stokes axis to get it down to 3 axes which is what our other one has
aux_img_cs = ia.coordsys()
# how to trim the freq axis--
#img_shape = (ia.shape())[0:3]
aux_img_shape = ia.shape()
print aux_img_shape
auxvals = np.squeeze(ia.getchunk())
auxmask = np.squeeze(ia.getchunk(getmask=True))
ia.close()
# get beam info
aux_hdr = imhead(imagename=aux_img, mode='summary')
#aux_bmaj_str = str(aux_hdr['restoringbeam']['major']['value'] * fudge_factor)+aux_hdr['restoringbeam']['major']['unit']
#aux_bmin_str = str(aux_hdr['restoringbeam']['minor']['value'] * fudge_factor)+aux_hdr['restoringbeam']['minor']['unit']
#aux_bpa_str = str(aux_hdr['restoringbeam']['positionangle']['value'])+aux_hdr['restoringbeam']['positionangle']['unit']
#aux_beam_fwhm = ( aux_hdr['restoringbeam']['major']['value'] * aux_hdr['restoringbeam']['minor']['value'] )**0.5
#print beam_fwhm
# create joint mask and zero out fluxes in either
# that are not in the joint mask
totmask = immask & auxmask
if flipMask:
totmask = ~totmask
imvals[~totmask] = 0.0
auxvals[~totmask] = 0.0
if dumbTest:
return imvals, auxvals, immask, auxmask
# size of peak search region in pixels -
pix_search_size = search_size * (beam_fwhm / mean_cell)
print "PIX SEARCH SIZE"
print pix_search_size
peak_indices = find_peaks(inimg=main_img,
search_size=pix_search_size,
thresh=thresh)
print "PIX INDICES"
print peak_indices
# radii in pixels
radii = np.arange(minrad, maxrad, radstep) * beam_fwhm / mean_cell
flux_vs_rad = np.array([])
rms_vs_rad = np.array([])
for i in radii:
# these return an array of fluxes of individual regions
mainInner, xx, yy, main_inner_count = sum_region_fluxes(imvals,
peak_indices,
radius=i)
auxInner, dumx, dumy, aux_inner_count = sum_region_fluxes(auxvals,
peak_indices,
radius=i)
if (doAnnuli):
mainOuter, xx, yy, main_outer_count = sum_region_fluxes(
imvals,
peak_indices,
radius=i + np.round(beam_fwhm / mean_cell))
auxOuter, dumx, dumy, aux_outer_count = sum_region_fluxes(
auxvals,
peak_indices,
radius=i + np.round(beam_fwhm / mean_cell))
main_flux = 2.0 * mainInner - mainOuter * main_inner_count / (
2 * main_inner_count - main_outer_count)
aux_flux = 2.0 * auxInner - auxOuter * aux_inner_count / (
2 * aux_inner_count - aux_outer_count)
else:
main_flux = mainInner
aux_flux = auxInner
# take the median and MAD of these
flux_vs_rad = np.append(flux_vs_rad, np.median(main_flux / aux_flux))
rms_vs_rad = np.append(rms_vs_rad, au.MAD(main_flux / aux_flux))
return radii * mean_cell, flux_vs_rad, rms_vs_rad, xx, yy, imvals
def compare_fluxes(img1="./ngvla_30dor/ngvla_30dor.ngvla-core-revC.autoDev2.image",img2="./ngvla_30dor/ngvla_30dor.ngvla-core-revC.skymodel.flat",mask1="",mask2="",search_size = 2.0, thresh = 4.5,search_img=1):
"""
img1,img2: casa images to compare
search_size: width of search region in beams (beam is takem from chosen search image)
thresh: threshold for peak detection (robust sigma above median, default = 4.5)
search_img: which image (1 or 2) to search for peaks
rstart, rstop, rdelta: start, stop, and delta radii for flux sums (in beams, again taken from search image)
images are presumed to be on the same grid to start with (shape, pixel size) -- use imgregrid task
and presumed to be in Jy/pixel (or otherwise same brightness unit per pixel) -- use combut.jyBm2jyPix()
returns: flux(search), flux(other), peak_locations
22jan2020 currently uses a square aperture (needs updating to circular) and hard wires the range of radii. return radii are in
pixels not arcsec.
"""
print "THRESH ", thresh
if (search_img == 1):
main_img = img1
aux_img = img2
else:
main_img = img2
aux_img = img1
print " main image = ",main_img
ia=iatool()
# main image characteristics
ia.open(main_img)
# average over the stokes axis to get it down to 3 axes which is what our other one has
main_img_cs = ia.coordsys()
# how to trim the freq axis--
#img_shape = (ia.shape())[0:3]
main_img_shape = ia.shape()
print main_img_shape
imvals=np.squeeze(ia.getchunk())
immask=np.squeeze(ia.getchunk(getmask=True))
ia.close()
# get beam info
hdr = imhead(imagename=main_img,mode='summary')
bmaj_str = str(hdr['restoringbeam']['major']['value'] )+hdr['restoringbeam']['major']['unit']
bmin_str = str(hdr['restoringbeam']['minor']['value'] )+hdr['restoringbeam']['minor']['unit']
bpa_str = str(hdr['restoringbeam']['positionangle']['value'])+hdr['restoringbeam']['positionangle']['unit']
beam_fwhm = ( hdr['restoringbeam']['major']['value'] * hdr['restoringbeam']['minor']['value'] )**0.5
print beam_fwhm
# secondary image characteristics
ia.open(aux_img)
# average over the stokes axis to get it down to 3 axes which is what our other one has
aux_img_cs = ia.coordsys()
# how to trim the freq axis--
#img_shape = (ia.shape())[0:3]
aux_img_shape = ia.shape()
print aux_img_shape
auxvals=np.squeeze(ia.getchunk())
auxmask=np.squeeze(ia.getchunk(getmask=True))
ia.close()
# get beam info
aux_hdr = imhead(imagename=aux_img,mode='summary')
#aux_bmaj_str = str(aux_hdr['restoringbeam']['major']['value'] * fudge_factor)+aux_hdr['restoringbeam']['major']['unit']
#aux_bmin_str = str(aux_hdr['restoringbeam']['minor']['value'] * fudge_factor)+aux_hdr['restoringbeam']['minor']['unit']
#aux_bpa_str = str(aux_hdr['restoringbeam']['positionangle']['value'])+aux_hdr['restoringbeam']['positionangle']['unit']
#aux_beam_fwhm = ( aux_hdr['restoringbeam']['major']['value'] * aux_hdr['restoringbeam']['minor']['value'] )**0.5
#print beam_fwhm
# create joint mask and zero out fluxes in either
# that are not in the joint mask
totmask = immask & auxmask
imvals[~totmask] = 0.0
auxvals[~totmask]=0.0
# set this correctly -
pix_search_size = search_size * 8.0
peak_indices = find_peaks(inimg=main_img,search_size = pix_search_size, thresh = thresh)
# hard wire radii - geez dude! ---
radii = np.arange(500)*2.0 + 2.0
flux_vs_rad = np.array([])
rms_vs_rad = np.array([])
for i in radii:
# these return an array of fluxes of individual regions
main_flux,xx,yy = sum_region_fluxes(imvals,peak_indices,radius=i)
aux_flux, dumx,dumy = sum_region_fluxes(auxvals,peak_indices,radius=i)
# take the median and MAD of these
flux_vs_rad = np.append(flux_vs_rad,np.median(main_flux/aux_flux))
rms_vs_rad = np.append(rms_vs_rad,au.MAD(main_flux/aux_flux))
return radii,flux_vs_rad,rms_vs_rad