def get_gaussian_psf_template_3_5_20(pixel_fwhm=3., nbin=5):
nc = nbin**2
psfnew = Gaussian2DKernel((pixel_fwhm / 2.355) * nbin,
x_size=125,
y_size=125).array.astype(np.float32)
cf = psf_poly_fit(psfnew, nbin=nbin)
return psfnew, cf, nc, nbin
def get_psf_and_vals(path):
f = open(path)
psf = np.loadtxt(path, skiprows=1).astype(np.float32)
nc, nbin = [np.int32(i) for i in f.readline().split()]
f.close()
cf = psf_poly_fit(psf, nbin=nbin)
return psf, nc, cf
def get_gaussian_psf_template_3_5_20(pixel_fwhm=3., nbin=5):
'''
Computes Gaussian PSF kernel for fast model evaluation with lion
Parameters
----------
pixel_fwhm : float
Full width at half-maximum (FWHM) of PSF in units of pixels.
Default is 3 (for SPIRE analysis).
nbin : int
Upsampling factor for sub-pixel interpolation method in lion
Default is 5.
Returns
-------
psfnew : '~numpy.ndarray'
Sum normalized PSF template up sampled by factor nbin in both x and y
cf : '~numpy.ndarray'
Coefficients of polynomial fit to upsampled PSF. These are used by lion
nbin : int
Upsampling factor, same as before..
'''
nc = nbin**2
psfnew = Gaussian2DKernel((pixel_fwhm / 2.355) * nbin,
x_size=125,
y_size=125).array.astype(np.float32)
cf = psf_poly_fit(psfnew, nbin=nbin)
return psfnew, cf, nc, nbin
def mock_pix_psf_files(self, two_star_blend, two_star_mode):
f = open(self.base_path + '/Data/sdss.0921/sdss.0921_psf.txt')
nc, nbin = [np.int32(i) for i in f.readline().split()]
f.close()
psf = np.loadtxt(self.base_path + '/Data/sdss.0921/sdss.0921_psf.txt',
skiprows=1).astype(np.float32)
cf = psf_poly_fit(psf, nbin=nbin)
nb = self.nbands
if two_star_blend:
if two_star_mode == 'rx3':
nb = 1
for b in xrange(nb):
f = open(
self.data_path + '/pixs/' + self.dir_name + '-pix' +
self.bands[b] + '.txt', 'w')
if two_star_blend:
f.write('%1d\t%1d\t1\n0.\t%0.3f\n%0.8f\t%1d' %
(self.imsz[0], self.imsz[1], self.gain,
self.nmgy_to_cts, self.truebacks[b]))
else:
f.write('%1d\t%1d\t1\n0.\t%0.3f\n%0.8f\t%1d' %
(self.imsz[0], self.imsz[1], self.gain,
self.nmgy_to_cts, self.normal_backs[b]))
np.savetxt(self.data_path + '/psfs/' + self.dir_name + '-psf' +
self.bands[b] + '.txt',
psf,
header='%1d\t%1d' % (nc, nbin),
comments='')
return nc, nbin, psf, cf
def get_gaussian_psf_template(fwhm, pixel_fwhm=3., nbin=5):
nc = nbin**2
psfnew = Gaussian2DKernel(pixel_fwhm / 2.355 * nbin,
x_size=125,
y_size=125).array.astype(np.float32)
psfnew2 = psfnew / np.max(psfnew) * nc
cf = psf_poly_fit(psfnew, nbin=nbin)
return psfnew2, cf, nc, nbin
def get_gaussian_psf_template(pixel_fwhm=3., nbin=5, normalization='max'):
nc = 25
psfnew = Gaussian2DKernel((pixel_fwhm / 2.355) * nbin,
x_size=125,
y_size=125).array.astype(np.float32)
if normalization == 'max':
print('Normalizing PSF by kernel maximum')
psfnew /= np.max(psfnew)
psfnew /= 4 * np.pi * (pixel_fwhm / 2.355)**2
else:
print('Normalizing PSF by kernel sum')
psfnew *= nc
cf = psf_poly_fit(psfnew, nbin=nbin)
return psfnew, cf, nc, nbin
def get_psf(self,
ifield=4,
band=0,
nx=1024,
ny=1024,
multfac=7.0,
nbin=0.,
poly_fit=True,
nwide=17,
width_psf_temp=50):
beta, rc, norm = find_psf_params(self.ciberdir + '/data/psfparams.txt',
tm=band + 1,
field=self.ciber_field_dict[ifield])
Nlarge = nx + 30 + 30
radmap = make_radius_map(
2 * Nlarge + nbin, 2 * Nlarge + nbin, Nlarge + nbin, Nlarge + nbin,
rc) * multfac # is the input supposed to be 2d?
self.psf_full = norm * np.power(1 + radmap, -3 * beta / 2.)
self.psf_full /= np.sum(self.psf_full)
self.psf_template = self.psf_full[Nlarge - nwide:Nlarge + nwide + 1,
Nlarge - nwide:Nlarge + nwide + 1]
# self.psf_template = self.psf_full[Nlarge-nwide:Nlarge+nwide, Nlarge-nwide:Nlarge+nwide]
print('imap center has shape', self.psf_template.shape)
if poly_fit:
psf = np.zeros((70, 70))
psf[0:35, 0:35] = self.psf_template
psf = np.array(
Image.fromarray(psf).resize((350, 350),
resample=Image.LANCZOS))
# psf = scipy.misc.imresize(psf, (350, 350), interp='lanczos', mode='F') # deprecated for Python 3
psfnew = np.array(psf[0:175, 0:175])
psfnew[0:173, 0:173] = psf[2:175,
2:175] # shift due to lanczos kernel
psfnew[0:173, 0:173] = psf[2:175,
2:175] # shift due to lanczos kernel
self.cf = psf_poly_fit(psfnew, nbin=5)
return beta, rc, norm
import os
from image_eval import image_model_eval, psf_poly_fit
from galaxy import retr_sers, retr_tranphon, to_moments
pixel_scale = 0.2 # arcsec
sigma_psf = 0.297 # arcsec, fwhm=0.7/2.355
sigma_in_pix = sigma_psf / pixel_scale
nc = 25 # PSF postage stamp size
nbin = 5 # upsampling
x = np.arange(-(nc / 2), (nc / 2) + 1, 1 / float(nbin))
r2 = x[:, None] * x[:, None] + x[None, :] * x[None, :]
psf = np.exp(-r2 /
(2. * sigma_in_pix * sigma_in_pix)) / (2 * np.pi * sigma_in_pix *
sigma_in_pix)
cf = psf_poly_fit(psf, nbin=nbin)
npar = cf.shape[0]
types = np.loadtxt('Data/phoSim_example.txt',
skiprows=9,
usecols=12,
dtype='string')
stars = (types == 'point')
galxs = (types == 'sersic2d')
mags = np.loadtxt('Data/phoSim_example.txt',
skiprows=9,
usecols=4,
dtype=np.float32)
star_mags = mags[stars]
galx_mags = mags[galxs]
rads = np.loadtxt('Data/phoSim_example.txt',
def writ_truedata():
pathlion = os.environ["LION_PATH"]
sys.path.insert(0, pathlion)
from image_eval import image_model_eval, psf_poly_fit
dictglob = dict()
sersindx = 2.
numbsidegrid = 20
pathliondata = os.environ["LION_DATA_PATH"] + '/data/'
pathlionimag = os.environ["LION_DATA_PATH"] + '/imag/'
fileobjt = open(pathliondata + 'sdss.0921_psf.txt')
numbsidepsfn, factsamp = [np.int32(i) for i in fileobjt.readline().split()]
fileobjt.close()
psfn = np.loadtxt(pathliondata + 'sdss.0921_psf.txt',
skiprows=1).astype(np.float32)
cpsf = psf_poly_fit(psfn, factsamp)
cpsf = cpsf.reshape((-1, cpsf.shape[2]))
np.random.seed(0)
numbside = [100, 100]
# generate stars
numbstar = 1
fluxdistslop = np.float32(2.0)
minmflux = np.float32(250.)
logtflux = np.random.exponential(scale=1. / (fluxdistslop - 1.),
size=numbstar).astype(np.float32)
dictglob['numbstar'] = numbstar
dictglob['xposstar'] = (np.random.uniform(size=numbstar) *
(numbside[0] - 1)).astype(np.float32)
dictglob['yposstar'] = (np.random.uniform(size=numbstar) *
(numbside[1] - 1)).astype(np.float32)
dictglob['fluxdistslop'] = fluxdistslop
dictglob['minmflux'] = minmflux
dictglob['fluxstar'] = minmflux * np.exp(logtflux)
dictglob['specstar'] = dictglob['fluxstar'][None, :]
dictglob['back'] = np.float32(179.)
dictglob['gain'] = np.float32(4.62)
# generate galaxies
numbgalx = 100
dictglob['numbgalx'] = numbgalx
dictglob['xposgalx'] = (np.random.uniform(size=numbgalx) *
(numbside[0] - 1)).astype(np.float32)
dictglob['yposgalx'] = (np.random.uniform(size=numbgalx) *
(numbside[1] - 1)).astype(np.float32)
dictglob['sizegalx'] = samp_powr(2., 10., 1.5, size=numbgalx)
dictglob['avecfrst'] = (2. * np.random.uniform(size=numbgalx) - 1.).astype(
np.float32)
dictglob['avecseco'] = (2. * np.random.uniform(size=numbgalx) - 1.).astype(
np.float32)
dictglob['avecthrd'] = (2. * np.random.uniform(size=numbgalx) - 1.).astype(
np.float32)
mgtd = np.sqrt(dictglob['avecfrst']**2 + dictglob['avecseco']**2 +
dictglob['avecthrd']**2)
dictglob['avecfrst'] /= mgtd
dictglob['avecseco'] /= mgtd
dictglob['avecthrd'] /= mgtd
dictglob['specgalx'] = samp_powr(2500., 25000., 2., size=numbgalx)[None, :]
dictglob['sbrtgalx'] = retr_sbrt(dictglob['specgalx'],
dictglob['sizegalx'], sersindx)
gdat = gdatstrt()
listsersindx = [0.5, 1., 2., 4., 6., 8., 10.]
for sersindx in listsersindx:
gdat.sersindx = sersindx
gridphon, amplphon = retr_sers(numbsidegrid=numbsidegrid,
sersindx=sersindx,
pathplot=pathlionimag)
gridphon, amplphon = retr_sers(numbsidegrid=numbsidegrid,
sersindx=sersindx,
pathplot=pathlionimag)
numbphongalx = (numbsidegrid + 1)**2
numbener = 1
numbphon = numbgalx * numbphongalx
xposphon = np.empty(numbphon)
yposphon = np.empty(numbphon)
specphon = np.empty((numbener, numbphon))
for k in range(dictglob['numbgalx']):
indx = np.arange(k * numbphongalx, (k + 1) * numbphongalx)
gridphontemp = gridphon * dictglob['sizegalx'][k]
xposphon[indx], yposphon[indx], specphon[:, indx] = retr_tranphon(gridphontemp, amplphon, dictglob['xposgalx'][k], dictglob['yposgalx'][k], dictglob['specgalx'][:, k], \
dictglob['avecfrst'][k], dictglob['avecseco'][k], dictglob['avecthrd'][k], 'fris')
xposcand = np.concatenate(
(xposphon, dictglob['xposstar'])).astype(np.float32)
yposcand = np.concatenate(
(yposphon, dictglob['yposstar'])).astype(np.float32)
speccand = np.concatenate((specphon, dictglob['specstar']),
axis=1).astype(np.float32)
indx = np.where((xposcand < 100.) & (xposcand > 0.) & (yposcand < 100.)
& (yposcand > 0.))[0]
xposcand = xposcand[indx]
yposcand = yposcand[indx]
speccand = speccand[:, indx]
# generate data
datacnts = image_model_eval(xposcand, yposcand, speccand[0, :],
dictglob['back'], numbside, numbsidepsfn,
cpsf)
#datacnts[datacnts < 1] = 1. # maybe some negative pixels
variance = datacnts / dictglob['gain']
datacnts += (np.sqrt(variance) *
np.random.normal(size=(numbside[1], numbside[0]))).astype(
np.float32)
dictglob['datacnts'] = datacnts
# auxiliary data
dictglob['numbside'] = numbside
dictglob['psfn'] = psfn
# write data to disk
path = pathliondata + 'true.h5'
filearry = h5py.File(path, 'w')
for attr, valu in dictglob.iteritems():
filearry.create_dataset(attr, data=valu)
filearry.close()
gdat.pathlionimag = pathlionimag
plot_cntsmaps(gdat, datacnts, 'truedatacnts')
plot_cntsmaps(gdat,
datacnts,
'truedatacntsphon',
xposcand=xposcand,
yposcand=yposcand,
speccand=speccand)
