def psf_ave(psfs_list,
not_count=(),
mode='CI',
mask_list=['default.reg'],
mask_img_list=None):
'''
Produce the average for a list of psfs.
Parameter
--------
psfs_list:
A raw list of psfs array.
mode: the way to do the average.
'direcity':
directly do the average of the scaled PSF
'CI':
Consider center (1/3 region) Intensity of the PSF. (weighted by
the root-square of the intensity),---as the noise of PSF is most
related to the Poission noise.The SNR of the image is related to
root-square of the image.
'CI_tot'
Silimar to CI, but use the total flux.
'mid'
Select the median value. But, four PSFs were recommed to load.
no_count:
The serial No. of psf which not considered.
mask_list:
A list include all the mask.reg names
Return
--------
A averaged and scaled PSF.
'''
### masked PSF give the region.
psf_NO = len(psfs_list)
for i in range(psf_NO):
if psfs_list[i] is None:
psfs_list[i] = np.zeros_like(
[x for x in psfs_list if x is not None][0])
psfs_l_msk = np.ones_like(psfs_list) # To load the PSF plus masked area
for i in range(psf_NO):
if i in not_count:
print("The PSF{0} is not count".format(i))
psfs_l_msk[i] = np.zeros_like(psfs_list[i])
else:
if mask_img_list is None:
msk_counts, mask_lists = text_in_string_list(
"PSF{0}_".format(i), mask_list)
mask = np.ones(psfs_list[i].shape)
if msk_counts != 0:
for j in range(msk_counts):
mask *= cr_mask(image=psfs_list[i],
filename=mask_lists[j])
else:
mask = mask_img_list[i]
psfs_l_msk[i] = psfs_list[i] * mask
# for i in range(psf_NO):
# print("plot psfs_list", i)
# plt.matshow(psfs_l_msk[i], origin= 'low', norm=LogNorm())
# plt.colorbar()
# plt.show()
### Doing the average.
if mode == 'direct':
for i in range(psf_NO):
if psfs_l_msk[i].sum() != 0:
psfs_l_msk[i] /= psfs_l_msk[i].sum(
) # scale the image to a same level
# print(np.where(np.isclose(psfs_l_msk,0)))
psf_ave = np.nanmean(np.where(np.isclose(psfs_l_msk, 0), np.nan,
psfs_l_msk),
axis=0)
# print(psf_ave.sum())
psf_std = np.nanstd(np.where(np.isclose(psfs_l_msk, 0), np.nan,
psfs_l_msk),
axis=0)
psf_std /= psf_ave.sum()
psf_ave /= psf_ave.sum()
elif mode == 'CI':
weights = np.zeros(psf_NO)
for i in range(psf_NO):
box_c = len(psfs_l_msk[i]) / 2
box_r = len(psfs_l_msk[i]) / 6
if psfs_l_msk[i].sum() != 0:
weights[i] = np.sqrt(
np.sum(psfs_l_msk[i][
box_c - box_r:box_c + box_r, box_c - box_r:box_c +
box_r])) # set weight based on their intensity (SNR)
print(
"Sum flux for PSF center", i, ":",
psfs_l_msk[i][box_c - box_r:box_c + box_r,
box_c - box_r:box_c + box_r].sum())
psfs_l_msk[i] /= weights[
i]**2 # scale the image to a same level
print("The final weights for doing the average:\n", weights)
# print(abs(psfs_l_msk[3]).min())
psfs_msk2nan = np.where(
np.isclose(psfs_l_msk, 0, rtol=1e-10, atol=1e-09), np.nan,
psfs_l_msk)
cleaned_psfs = np.ma.masked_array(psfs_msk2nan, np.isnan(psfs_msk2nan))
# plt.imshow(cleaned_psfs.mask[5],origin = 'low')
# plt.show()
psf_ave = np.ma.average(cleaned_psfs, axis=0, weights=weights)
diffs = (psfs_l_msk - psf_ave)**2
diffs_msk2nan = np.where(
np.isclose(psfs_l_msk, 0, rtol=1e-10, atol=1e-09), np.nan, diffs)
cleaned_diffs = np.ma.masked_array(diffs_msk2nan,
np.isnan(diffs_msk2nan))
psf_variance = np.ma.average(cleaned_diffs, weights=weights, axis=0)
psf_std = np.sqrt(psf_variance)
psf_std /= psf_ave.sum()
psf_ave /= psf_ave.sum()
psf_std = psf_std.data
psf_ave = psf_ave.data
elif mode == 'CI_tot':
weights = np.zeros(psf_NO)
for i in range(psf_NO):
if psfs_l_msk[i].sum() != 0:
weights[i] = np.sqrt(
np.sum(psfs_l_msk[i]
)) # set weight based on their intensity (SNR)
print("Sum flux for PSF", i, ":", psfs_l_msk[i].sum())
psfs_l_msk[i] /= psfs_l_msk[i].sum(
) # scale the image to a same level
print("The final weights for doing the average:\n", weights)
# print(abs(psfs_l_msk[3]).min())
psfs_msk2nan = np.where(
np.isclose(psfs_l_msk, 0, rtol=1e-10, atol=1e-09), np.nan,
psfs_l_msk)
cleaned_psfs = np.ma.masked_array(psfs_msk2nan, np.isnan(psfs_msk2nan))
# plt.imshow(cleaned_psfs.mask[5],origin = 'low')
# plt.show()
psf_ave = np.ma.average(cleaned_psfs, axis=0, weights=weights)
diffs = (psfs_l_msk - psf_ave)**2
diffs_msk2nan = np.where(
np.isclose(psfs_l_msk, 0, rtol=1e-10, atol=1e-09), np.nan, diffs)
cleaned_diffs = np.ma.masked_array(diffs_msk2nan,
np.isnan(diffs_msk2nan))
psf_variance = np.ma.average(cleaned_diffs, weights=weights, axis=0)
psf_std = np.sqrt(psf_variance)
psf_std /= psf_ave.sum()
psf_ave /= psf_ave.sum()
psf_std = psf_std.data
psf_ave = psf_ave.data
elif mode == 'mid':
weights = np.zeros(psf_NO)
for i in range(psf_NO):
box_c = len(psfs_l_msk[i]) / 2
box_r = len(psfs_l_msk[i]) / 6
if psfs_l_msk[i].sum() != 0:
weights[i] = np.sqrt(
np.sum(psfs_l_msk[i][
box_c - box_r:box_c + box_r, box_c - box_r:box_c +
box_r])) # set weight based on their intensity (SNR)
psfs_l_msk[i] /= weights[
i]**2 # scale the image to a same level
sz = len(psfs_l_msk[0])
psf_ave = np.zeros_like(psfs_l_msk[0])
for i in range(sz):
for j in range(sz):
psf_cell = psfs_l_msk[:, i, j]
psf_cell = psf_cell[psf_cell != 0.] # Delete the non-zeros.
psf_ave[i, j] = median(psf_cell)
psf_ave /= psf_ave.sum()
#### The PSF are found not very necessary to be shiftted. !!!! Note the high_CI is not ready --- high_res. mask is not OK.
# if mode == 'high_CI':
# psfs_high_list = np.empty([psf_NO, psfs_list[0].shape[0]*scale, psfs_list[0].shape[1]*scale])
# #Creat a mask_high_list:
# mask_high_list = np.ones_like(psfs_high_list)
# for i in range(psf_NO):
# psfs_high_list[i] = im_2_high_res(psfs_list[i], scale=scale) # scale the image to a same level
# psfs_high_list[i] *= mask_high_list[i]
# psfs_high_list[i] /= np.sqrt(np.sum(psfs_high_list[0]))
# psf_high_total = np.sum(psfs_high_list, axis=0)
# sum_4ave = np.sum(mask_high_list, axis=0)
# print(sum_4ave)
# psf_high_final = psf_high_total/sum_4ave
# psf_final = rebin(psf_high_final, scale = scale)
else:
raise ValueError("mode is not defined")
if mode == 'mid':
return psf_ave
else:
return psf_ave, psf_std
kwargs_lens_list[0]['center_y'] += 0.08125
kwargs_lens_light_list[0]['center_x'] += 0.08125
kwargs_lens_light_list[0]['center_y'] += 0.08125
kwargs_source_list[0]['center_x'] += 0.08125
kwargs_source_list[0]['center_y'] += 0.08125
kwargs_constraints = {}
if glob.glob(folder + savename) == []:
# raise ValueError("The first time run of with QSO case is not finished")
# else:
# #Load the result from the first run:
# _, _, kwargs_result, _, _, _ = pickle.load(open(qso_folder+'model_result.pkl','rb'))
# fixed_lens, fixed_source, fixed_lens_light, fixed_ps, fixed_cosmo = fix_setting
#Setting up the fitting:
lens_data = pyfits.getdata(folder + 'non_drizzled-image-1.fits')
len_std = pyfits.getdata(folder + 'non_drizzled-noise_map-1.fits')
lens_mask = cr_mask(lens_data, 'normal_nondrz_mask.reg')
framesize = 57
ct = int((len(lens_data) - framesize) / 2)
lens_data = lens_data[ct:-ct, ct:-ct]
len_std = len_std[ct:-ct, ct:-ct]
lens_mask = (1 - lens_mask)[ct:-ct, ct:-ct]
plt.imshow(lens_data, origin='lower', cmap='gist_heat', norm=LogNorm())
plt.colorbar()
plt.show()
# exp_time = 599.* 2 * 8
# stdd = 0.016/np.sqrt(8) #Measurement from empty retion, 0.016*0.08**2/0.13**2/np.sqrt(8)
# len_std = (abs(lens_data/exp_time)+stdd**2)**0.5
deltaPix = 0.13
psf = pyfits.getdata(folder + 'non_drizzled_psf-1.fits')
psf_fsize = 47
'num_point_source_list': [len(kwargs_ps['ra_image'])],
'solver_type':
solver_type, # 'PROFILE', 'PROFILE_SHEAR', 'ELLIPSE', 'CENTER'
'Ddt_sampling': True,
}
if glob.glob(qso_folder + with_qso_savename) == []:
raise ValueError("The first time run of with QSO case is not finished")
if glob.glob(folder + savename) == []:
#Load the result from the first run:
_, _, kwargs_result, _, _, _ = pickle.load(
open(qso_folder + with_qso_savename, 'rb'))
# fixed_lens, fixed_source, fixed_lens_light, fixed_ps, fixed_cosmo = fix_setting
#Setting up the fitting:
lens_data = pyfits.getdata(folder + 'Drz_QSO_image.fits')
# len_std = pyfits.getdata(folder+'noise_map.fits')
lens_mask = cr_mask(lens_data, 'normal_mask.reg')
framesize = 81
ct = int((len(lens_data) - framesize) / 2)
lens_data = lens_data[ct:-ct, ct:-ct]
# len_std = len_std[ct:-ct,ct:-ct]
lens_mask = (1 - lens_mask)[ct:-ct, ct:-ct]
plt.imshow(lens_data, origin='lower', cmap='gist_heat', norm=LogNorm())
plt.colorbar()
plt.show()
exp_time = 599. * 2 * 8
stdd = 0.0008 #Measurement from empty retion, 0.016*0.08**2/0.13**2/np.sqrt(8)
# len_std = (abs(lens_data/exp_time)+stdd**2)**0.5
len_std = (abs(lens_data / exp_time) * 8 + stdd**2)**0.5
# vgrad = np.gradient(lens_data)
'num_point_source_list': [len(kwargs_ps['ra_image'])],
'solver_type':
solver_type, # 'PROFILE', 'PROFILE_SHEAR', 'ELLIPSE', 'CENTER'
'Ddt_sampling': True,
}
if glob.glob(qso_folder + with_qso_savename) == []:
raise ValueError("The first time run of with QSO case is not finished")
if glob.glob(folder + savename) == []:
#Load the result from the first run:
_, _, kwargs_result, _, _, _ = pickle.load(
open(qso_folder + with_qso_savename, 'rb'))
# fixed_lens, fixed_source, fixed_lens_light, fixed_ps, fixed_cosmo = fix_setting
#Setting up the fitting:
lens_data = pyfits.getdata(folder + 'Drz_QSO_image.fits')
# len_std = pyfits.getdata(folder+'noise_map.fits')
lens_mask = cr_mask(lens_data, 'big_mask.reg')
framesize = 195
ct = int((len(lens_data) - framesize) / 2)
lens_data = lens_data[ct:-ct, ct:-ct]
# len_std = len_std[ct:-ct,ct:-ct]
lens_mask = (1 - lens_mask)[ct:-ct, ct:-ct]
plt.imshow(lens_data * lens_mask,
origin='lower',
cmap='gist_heat',
norm=LogNorm())
plt.colorbar()
plt.show()
exp_time = 599. * 2 * 8
stdd = 0.0004 #Measurement from empty retion, 0.016*0.08**2/0.13**2/np.sqrt(8)
# len_std = (abs(lens_data/exp_time)+stdd**2)**0.5
