def fit_specindex(self, freqarr, fluxarr, efluxarr, do_log=False):
""" Fits spectral index to data.
do_log is True/False implies you fit spectral index in logFlux vs logFreq space or not."""
import functions as func
import math
from scipy.optimize import leastsq
x = freqarr
flux = fluxarr
eflux = efluxarr
f0 = N.median(x)
mask = N.zeros(len(fluxarr), dtype=bool)
if do_log:
x = N.log10(x/f0); y = N.log10(flux); sig = N.abs(eflux/flux)/2.303
funct = func.poly
else:
x = x/f0; y = flux; sig = eflux
funct = func.sp_in
spin, espin = func.fit_mask_1d(x, y, sig, mask, funct, do_err=True, order=1)
if do_log:
spin[0] = math.pow(10.0, spin[0])
espin[0] = spin[0]*math.log(10.0)*espin[0]
return spin[0], spin[1], espin[1]
def shapelet_getroot(xfn, yfn, xco, xcen, ycen):
""" This finds the root for finding the shapelet centre. If there are multiple roots, takes
that which closest to the 'centre', taken as the intensity barycentre. This is the python
version of getroot.f of anaamika."""
import functions as func
root = None
npoint = len(xfn)
error = 0
if npoint == 0:
error = 1
elif yfn.max() * yfn.min() >= 0.:
error = 1
minint = 0
minintold = 0
for i in range(1, npoint):
if yfn[i - 1] * yfn[i] < 0.:
if minintold == 0: # so take nearest to centre
if abs(yfn[i - 1]) < abs(yfn[i]):
minint = i - 1
else:
minint = i
else:
dnew = func.dist_2pt([xco, xfn[i]], [xcen, ycen])
dold = func.dist_2pt([xco, xfn[minintold]], [xcen, ycen])
if dnew <= dold:
minint = i
else:
minint = minintold
minintold = minint
if minint < 1 or minint > npoint: error = 1
if error != 1:
low = minint - min(2, minint) #-1)
up = minint + min(2,
npoint - 1 - minint) # python array indexing rubbish
nfit = up - low + 1
xfit = xfn[low:low + nfit]
yfit = yfn[low:low + nfit]
sig = N.ones(nfit)
smask = N.zeros(nfit, dtype=bool)
xx = [i for i in range(low, low + nfit)]
[c, m], errors = func.fit_mask_1d(xfit,
yfit,
sig,
smask,
func.poly,
do_err=False,
order=1)
root = -c / m
if root < xfn[low] or root > xfn[up]: error = 1
return root, error
def shapelet_getroot(xfn, yfn, xco, xcen, ycen):
""" This finds the root for finding the shapelet centre. If there are multiple roots, takes
that which closest to the 'centre', taken as the intensity barycentre. This is the python
version of getroot.f of anaamika."""
import functions as func
root = None
npoint = len(xfn)
error = 0
if npoint == 0:
error = 1
elif yfn.max() * yfn.min() >= 0.0:
error = 1
minint = 0
minintold = 0
for i in range(1, npoint):
if yfn[i - 1] * yfn[i] < 0.0:
if minintold == 0: # so take nearest to centre
if abs(yfn[i - 1]) < abs(yfn[i]):
minint = i - 1
else:
minint = i
else:
dnew = func.dist_2pt([xco, xfn[i]], [xcen, ycen])
dold = func.dist_2pt([xco, xfn[minintold]], [xcen, ycen])
if dnew <= dold:
minint = i
else:
minint = minintold
minintold = minint
if minint < 1 or minint > npoint:
error = 1
if error != 1:
low = minint - min(2, minint) # -1)
up = minint + min(2, npoint - 1 - minint) # python array indexing rubbish
nfit = up - low + 1
xfit = xfn[low : low + nfit]
yfit = yfn[low : low + nfit]
sig = N.ones(nfit)
smask = N.zeros(nfit, dtype=bool)
xx = [i for i in range(low, low + nfit)]
[c, m], errors = func.fit_mask_1d(xfit, yfit, sig, smask, func.poly, do_err=False, order=1)
root = -c / m
if root < xfn[low] or root > xfn[up]:
error = 1
return root, error
def shape_findcen(image, mask, basis, beta, nmax, beam_pix): # + check_cen_shapelet
""" Finds the optimal centre for shapelet decomposition. Minimising various
combinations of c12 and c21, as in literature doesnt work for all cases.
Hence, for the c1 image, we find the zero crossing for every vertical line
and for the c2 image, the zero crossing for every horizontal line, and then
we find intersection point of these two. This seems to work even for highly
non-gaussian cases. """
import functions as func
import sys
hc = []
hc = shapelet_coeff(nmax, basis)
msk = N.zeros(mask.shape, dtype=bool)
for i, v in N.ndenumerate(mask):
msk[i] = not v
n, m = image.shape
cf12 = N.zeros(image.shape, dtype=N.float32)
cf21 = N.zeros(image.shape, dtype=N.float32)
index = [(i, j) for i in range(n) for j in range(m)]
for coord in index:
if msk[coord]:
B12 = shapelet_image(basis, beta, coord, hc, 0, 1, image.shape)
cf12[coord] = N.sum(image * B12 * msk)
if coord == (27, 51):
dumpy = B12
B21 = shapelet_image(basis, beta, coord, hc, 1, 0, image.shape)
cf21[coord] = N.sum(image * B21 * msk)
else:
cf12[coord] = None
cf21[coord] = None
(xmax, ymax) = N.unravel_index(image.argmax(), image.shape) # FIX with mask
if xmax in [1, n] or ymax in [1, m]:
(m1, m2, m3) = func.moment(mask)
xmax, ymax = N.round(m2)
# in high snr area, get zero crossings for each horizontal and vertical line for c1, c2 resp
tr_mask = mask.transpose()
tr_cf21 = cf21.transpose()
try:
(x1, y1) = getzeroes_matrix(mask, cf12, ymax, xmax) # y1 is array of zero crossings
(y2, x2) = getzeroes_matrix(tr_mask, tr_cf21, xmax, ymax) # x2 is array of zero crossings
# find nominal intersection pt as integers
xind = N.where(x1 == xmax)
yind = N.where(y2 == ymax)
xind = xind[0][0]
yind = yind[0][0]
# now take 2 before and 2 after, fit straight lines, get proper intersection
ninter = 5
if xind < 3 or yind < 3 or xind > n - 2 or yind > m - 2:
ninter = 3
xft1 = x1[xind - (ninter - 1) / 2 : xind + (ninter - 1) / 2 + 1]
yft1 = y1[xind - (ninter - 1) / 2 : xind + (ninter - 1) / 2 + 1]
xft2 = x2[yind - (ninter - 1) / 2 : yind + (ninter - 1) / 2 + 1]
yft2 = y2[yind - (ninter - 1) / 2 : yind + (ninter - 1) / 2 + 1]
sig = N.ones(ninter, dtype=float)
smask1 = N.array([r == 0 for r in yft1])
smask2 = N.array([r == 0 for r in xft2])
cen = [0.0] * 2
if sum(smask1) < len(yft1) and sum(smask2) < len(xft2):
[c1, m1], errors = func.fit_mask_1d(xft1, yft1, sig, smask1, func.poly, do_err=False, order=1)
[c2, m2], errors = func.fit_mask_1d(xft2, yft2, sig, smask2, func.poly, do_err=False, order=1)
if m2 - m1 == 0:
cen[0] = cen[1] = 0.0
else:
cen[0] = (c1 - c2) / (m2 - m1)
cen[1] = c1 + m1 * cen[0]
else:
cen[0] = cen[1] = 0.0
# check if estimated centre makes sense
error = shapelet_check_centre(image, mask, cen, beam_pix)
except:
error = 1
if error > 0:
# print 'Error '+str(error)+' in finding centre, will take 1st moment instead.'
(m1, m2, m3) = func.moment(image, mask)
cen = m2
return cen
def shape_findcen(image, mask, basis, beta, nmax,
beam_pix): # + check_cen_shapelet
""" Finds the optimal centre for shapelet decomposition. Minimising various
combinations of c12 and c21, as in literature doesnt work for all cases.
Hence, for the c1 image, we find the zero crossing for every vertical line
and for the c2 image, the zero crossing for every horizontal line, and then
we find intersection point of these two. This seems to work even for highly
non-gaussian cases. """
import functions as func
import sys
hc = []
hc = shapelet_coeff(nmax, basis)
msk = N.zeros(mask.shape, dtype=bool)
for i, v in N.ndenumerate(mask):
msk[i] = not v
n, m = image.shape
cf12 = N.zeros(image.shape, dtype=N.float32)
cf21 = N.zeros(image.shape, dtype=N.float32)
index = [(i, j) for i in range(n) for j in range(m)]
for coord in index:
if msk[coord]:
B12 = shapelet_image(basis, beta, coord, hc, 0, 1, image.shape)
cf12[coord] = N.sum(image * B12 * msk)
if coord == (27, 51): dumpy = B12
B21 = shapelet_image(basis, beta, coord, hc, 1, 0, image.shape)
cf21[coord] = N.sum(image * B21 * msk)
else:
cf12[coord] = None
cf21[coord] = None
(xmax, ymax) = N.unravel_index(image.argmax(),
image.shape) # FIX with mask
if xmax in [1, n] or ymax in [1, m]:
(m1, m2, m3) = func.moment(mask)
xmax, ymax = N.round(m2)
# in high snr area, get zero crossings for each horizontal and vertical line for c1, c2 resp
tr_mask = mask.transpose()
tr_cf21 = cf21.transpose()
try:
(x1, y1) = getzeroes_matrix(mask, cf12, ymax,
xmax) # y1 is array of zero crossings
(y2, x2) = getzeroes_matrix(tr_mask, tr_cf21, xmax,
ymax) # x2 is array of zero crossings
# find nominal intersection pt as integers
xind = N.where(x1 == xmax)
yind = N.where(y2 == ymax)
xind = xind[0][0]
yind = yind[0][0]
# now take 2 before and 2 after, fit straight lines, get proper intersection
ninter = 5
if xind < 3 or yind < 3 or xind > n - 2 or yind > m - 2:
ninter = 3
xft1 = x1[xind - (ninter - 1) / 2:xind + (ninter - 1) / 2 + 1]
yft1 = y1[xind - (ninter - 1) / 2:xind + (ninter - 1) / 2 + 1]
xft2 = x2[yind - (ninter - 1) / 2:yind + (ninter - 1) / 2 + 1]
yft2 = y2[yind - (ninter - 1) / 2:yind + (ninter - 1) / 2 + 1]
sig = N.ones(ninter, dtype=float)
smask1 = N.array([r == 0 for r in yft1])
smask2 = N.array([r == 0 for r in xft2])
cen = [0.] * 2
if sum(smask1) < len(yft1) and sum(smask2) < len(xft2):
[c1, m1], errors = func.fit_mask_1d(xft1,
yft1,
sig,
smask1,
func.poly,
do_err=False,
order=1)
[c2, m2], errors = func.fit_mask_1d(xft2,
yft2,
sig,
smask2,
func.poly,
do_err=False,
order=1)
if m2 - m1 == 0:
cen[0] = cen[1] = 0.0
else:
cen[0] = (c1 - c2) / (m2 - m1)
cen[1] = c1 + m1 * cen[0]
else:
cen[0] = cen[1] = 0.0
# check if estimated centre makes sense
error = shapelet_check_centre(image, mask, cen, beam_pix)
except:
error = 1
if error > 0:
#print 'Error '+str(error)+' in finding centre, will take 1st moment instead.'
(m1, m2, m3) = func.moment(image, mask)
cen = m2
return cen
