"""
Utility functions.
Many of them are partially implemented. They were in the
old, non-modular version of PyKOALA and have not been included
in the current modular scheme.
"""
# =============================================================================
# Basics packages
# =============================================================================
import numpy as np
from matplotlib import pyplot as plt
from scipy import interpolate
from scipy import optimize
# =============================================================================
# Astropy and associated packages
# =============================================================================
from astropy.io import fits
# =============================================================================
# =============================================================================
# PyKOALA modules
# =============================================================================
from pykoala import vprint
# =============================================================================
# Ancillary Functions - RSS Related
# =============================================================================
[docs]def detect_edge(rss):
"""
Detect the edges of a RSS. Returns the minimum and maximum wavelength that
determine the maximum interval with valid (i.e. no masked) data in all the
spaxels.
Parameters
----------
rss : RSS object.
Returns
-------
min_w : float
The lowest value (in units of the RSS wavelength) with
valid data in all spaxels.
min_index : int
Index of min_w in the RSS wavelength variable.
max_w : float
The higher value (in units of the RSS wavelength) with
valid data in all spaxels.
max_index : int
Index of max_w in the RSS wavelength variable.
"""
collapsed = np.sum(rss.intensity, 0)
nans = np.isfinite(collapsed)
wavelength = rss.wavelength
min_w = wavelength[nans].min()
min_index = wavelength.tolist().index(min_w)
max_w = wavelength[nans].max()
max_index = wavelength.tolist().index(max_w)
return min_w, min_index, max_w, max_index
# ----------------------------------------------------------------------------------------------------------------------
# WCS operations
# ----------------------------------------------------------------------------------------------------------------------
[docs]def update_wcs_coords(wcs, ra_dec_val=None, ra_dec_offset=None):
"""Update the celestial reference values of a WCS.
Update the celestial coordinates of a WCS using new central values of RA
and DEC or relative offsets expressed in degree.
Parameters
----------
- wcs: asteropy.wcs.WCS
Target WCS to update.
- ra_dec_val: list or tupla, default=None
New CRVAL of RA and DEC.
- ra_dec_offset: list or tupla, default=None
Relative offset that will be applyied to CRVAL of RA and DEC axis. If
`ra_dec_val` is privided, this will be ignored.
Return
------
- correct_wcs: astropy.wcs.WCS
A copy of the original WCS with the reference values updated.
"""
correc_wcs = wcs.deepcopy()
if ra_dec_val is not None:
if "RA" in correc_wcs.wcs.ctype[0]:
correc_wcs.wcs.crval[0] = ra_dec_val[0]
correc_wcs.wcs.crval[1] = ra_dec_val[1]
elif "RA" in correc_wcs.wcs.ctype[1]:
correc_wcs.wcs.crval[0] = ra_dec_val[1]
correc_wcs.wcs.crval[1] = ra_dec_val[0]
else:
raise NameError(
"RA coordinate could not be found in the WCS coordinate types:"
+ f"{correc_wcs.wcs.ctype[0]}, {correc_wcs.wcs.ctype[1]}")
elif ra_dec_offset is not None:
if "RA" in correc_wcs.wcs.ctype[0]:
correc_wcs.wcs.crval[0] = correc_wcs.wcs.crval[0] + \
ra_dec_offset[0]
correc_wcs.wcs.crval[1] = correc_wcs.wcs.crval[1] + \
ra_dec_offset[1]
elif "RA" in correc_wcs.wcs.ctype[1]:
correc_wcs.wcs.crval[0] = correc_wcs.wcs.crval[0] + \
ra_dec_offset[1]
correc_wcs.wcs.crval[1] = correc_wcs.wcs.crval[1] + \
ra_dec_offset[0]
else:
raise NameError(
"RA coordinate could not be found in the WCS coordinate types:"
+ f"{correc_wcs.wcs.ctype[0]}, {correc_wcs.wcs.ctype[1]}")
return correc_wcs
# ----------------------------------------------------------------------------------------------------------------------
# Arithmetic operations
# ----------------------------------------------------------------------------------------------------------------------
[docs]def med_abs_dev(x, axis=0):
mad = np.nanmedian(
np.abs(x - np.expand_dims(np.nanmedian(x, axis=axis), axis=axis)),
axis=axis)
return mad
[docs]def std_from_mad(x, axis=0, k=1.4826):
mad = med_abs_dev(x, axis=axis)
return k * mad
[docs]def running_mean(x, n_window):
"""
This function calculates the running mean of an array.
Parameters
----------
x : (n,) np.ndarray
This is the given array.
n_window : Int
Number of neighbours for computing the running mean.
Returns
-------
running_mean_x: (n - n_window,) np.ndarray
Running mean array.
"""
cumsum = np.cumsum(np.insert(x, 0, 0))
return (cumsum[n_window:] - cumsum[:-n_window]) / n_window
[docs]def flux_conserving_interpolation(new_wavelength, wavelength, spectra, **interp_args):
"""Flux-conserving linear interpolation.
Linear interpolation of a spectrum :math:`I_\lambda(\labmda)`
as a function of wavelength :math:`\lambda`,
ensuring that the integrated flux
math::
F(\lambda_a, \lambda_b)
= \int_{\lambda_a}^{\lambda_b} I_\lambda(\labmda) d\lambda$
is conserved for any :math:`(\lambda_a, \lambda_b)`.
`np.nan` values become zero.
Parameters
----------
new_wavelength : ndarray
New values of the x coordinate (wavelength) :math:`\lambda_{new}`.
wavelength : ndarray
Old values of the x coordinate (wavelength) :math:`\lambda`.
spectra : ndarray
Old values of the y coordinate (spectrum) :math:`I_\lambda(\labmda)`.
**interp_args : dict, optional
Additional parameters to be passed to `np.interp`
Returns
-------
ndarray
Interpolated spectra :math:`I_\lambda(\labmda_{new})`.
Notes
-----
The function computes the cumulative flux with `np.nancumsum`,
calls `np.interp`, and differentiates back.
"""
dwave = wavelength[1:] - wavelength[:-1]
wavelength_edges = np.hstack((wavelength[0] - dwave[0] / 2, wavelength[:-1] + dwave / 2,
wavelength[-1] + dwave[-1] / 2))
new_dwave = new_wavelength[1:] - new_wavelength[:-1]
new_wavelength_edges = np.hstack((new_wavelength[0] - new_dwave[0] / 2, new_wavelength[:-1] + new_dwave / 2,
new_wavelength[-1] + new_dwave[-1] / 2))
cum_spectra = np.nancumsum(np.diff(wavelength_edges) * spectra)
cum_spectra = np.hstack((0, cum_spectra))
new_cum_spectra = np.interp(
new_wavelength_edges, wavelength_edges, cum_spectra, **interp_args)
new_spectra = np.diff(new_cum_spectra) / np.diff(new_wavelength_edges)
return new_spectra
[docs]def centre_of_mass(w, x, y):
"""Compute the centre of mass of a given image.
Parameters
----------
w: np.ndarray(float)
(n,) weights computing the centre of mass.
x: np.ndarray(float)
(n,) Coordinates corresponding to the x-axis (columns).
y: np.ndarray(float)
(n,) Coordinates corresponding to the y-axis (rows).
Returns
-------
x_com: float
y_com: float
"""
norm = np.nansum(w)
x_com, y_com = np.nansum(w * x) / norm, np.nansum(w * y) / norm
if np.isfinite(x_com) and np.isfinite(y_com):
return x_com, y_com
else:
raise RuntimeError(
"Failed computing centre of mass computed for\n w={}\n x={}\n y={}"
.format(w, x, y))
[docs]def growth_curve_1d(f, x, y):
"""TODO"""
r2 = x**2 + y**2
idx_sorted = np.argsort(r2)
growth_c = np.nancumsum(f[idx_sorted])
return r2[idx_sorted], growth_c
[docs]def growth_curve_2d(image, x0=None, y0=None):
"""Compute the curve of growth of an array f with respect to a given point (x0, y0).
Parameters
----------
image: np.ndarray
2D array
x0: float
Origin of coordinates in pixels along the x-axis (columns).
y0: float
Origin of coordinates in pixels along the y-axis (rows).
Returns
-------
r2: np.ndarray
Vector containing the square radius with respect to (x0, y0).
growth_curve: np.ndarray
Curve of growth centered at (x0, y0).
"""
xx, yy = np.meshgrid(np.arange(0, image.shape[1]),
np.arange(0, image.shape[1]))
r2 = (xx - x0) ** 2 + (yy - y0) ** 2
idx_sorted = np.argsort(r2)
growth_c = np.cumsum(image.flatten()[idx_sorted])
return r2[idx_sorted], growth_c
[docs]def interpolate_image_nonfinite(image):
"""Use scipy.interpolate.NearestNDInterpolator to replace NaN values.
Parameters
----------
- image: (np.ndarray)
2D array to be interpolated
Returnrs
--------
- interpolated_image: (np.ndarray)
"""
if image.ndim != 2:
raise ArithmeticError(f"Input image must have 2D not {image.ndim}")
x, y = np.meshgrid(
np.arange(0, image.shape[1], 1),
np.arange(0, image.shape[0], 1))
mask = np.isfinite(image)
interp = interpolate.NearestNDInterpolator(
list(zip(x[mask], y[mask])), image[mask])
interp_image = interp(x, y)
return interp_image
# TODO: refactor
[docs]def smooth_spectrum(wlm, s, wave_min=0, wave_max=0, step=50, exclude_wlm=[[0, 0]], order=7,
weight_fit_median=0.5, plot=False, verbose=False, fig_size=12):
"""
THIS IS NOT EXACTLY THE SAME THING THAT applying signal.medfilter()
This needs to be checked, updated, and combine (if needed) with task fit_smooth_spectrum.
The task gets the median value in steps of "step", gets an interpolated spectrum,
and fits a 7-order polynomy.
It returns fit_median + fit_median_interpolated (each multiplied by their weights).
Tasks that use this: get_telluric_correction
"""
if verbose:
vprint("\n> Computing smooth spectrum...")
if wave_min == 0:
wave_min = wlm[0]
if wave_max == 0:
wave_max = wlm[-1]
running_wave = []
running_step_median = []
cuts = np.int((wave_max - wave_min) / step)
exclude = 0
corte_index = -1
for corte in range(cuts+1):
next_wave = wave_min+step*corte
if next_wave < wave_max:
if next_wave > exclude_wlm[exclude][0] and next_wave < exclude_wlm[exclude][1]:
if verbose:
vprint(f" Skipping {next_wave}"
+ f" as it is in the exclusion range [{exclude_wlm[exclude][0]}, {exclude_wlm[exclude][1]}]")
else:
corte_index = corte_index+1
running_wave.append(next_wave)
region = np.where(
(wlm > running_wave[corte_index]-step/2) & (wlm < running_wave[corte_index]+step/2))
running_step_median.append(np.nanmedian(s[region]))
if next_wave > exclude_wlm[exclude][1]:
exclude = exclude + 1
# if verbose and exclude_wlm[0] != [0,0] : print "--- End exclusion range ",exclude
if exclude == len(exclude_wlm):
exclude = len(exclude_wlm)-1
running_wave.append(wave_max)
region = np.where((wlm > wave_max-step) & (wlm < wave_max+0.1))
running_step_median.append(np.nanmedian(s[region]))
# Check not nan
_running_wave_ = []
_running_step_median_ = []
for i in range(len(running_wave)):
if np.isnan(running_step_median[i]):
if verbose:
vprint(f"There is a nan in {running_wave[i]}")
else:
_running_wave_.append(running_wave[i])
_running_step_median_.append(running_step_median[i])
fit = np.polyfit(_running_wave_, _running_step_median_, order)
pfit = np.poly1d(fit)
fit_median = pfit(wlm)
interpolated_continuum_smooth = interpolate.splrep(
_running_wave_, _running_step_median_, s=0.02)
fit_median_interpolated = interpolate.splev(
wlm, interpolated_continuum_smooth, der=0)
if plot:
plt.figure(figsize=(fig_size, fig_size/2.5))
plt.plot(wlm, s, alpha=0.5)
plt.plot(running_wave, running_step_median, "+", ms=15, mew=3)
plt.plot(wlm, fit_median, label="fit median")
plt.plot(wlm, fit_median_interpolated, label="fit median_interp")
plt.plot(wlm, weight_fit_median*fit_median + (1-weight_fit_median)
* fit_median_interpolated, label="weighted")
# extra_display = (np.nanmax(fit_median)-np.nanmin(fit_median)) / 10
# plt.ylim(np.nanmin(fit_median)-extra_display, np.nanmax(fit_median)+extra_display)
ymin = np.nanpercentile(s, 1)
ymax = np.nanpercentile(s, 99)
rango = (ymax-ymin)
ymin = ymin - rango/10.
ymax = ymax + rango/10.
plt.ylim(ymin, ymax)
plt.xlim(wlm[0]-10, wlm[-1]+10)
plt.minorticks_on()
plt.legend(frameon=False, loc=1, ncol=1)
plt.axvline(x=wave_min, color='k', linestyle='--')
plt.axvline(x=wave_max, color='k', linestyle='--')
plt.xlabel("Wavelength [$\mathrm{\AA}$]")
if exclude_wlm[0][0] != 0:
for i in range(len(exclude_wlm)):
plt.axvspan(exclude_wlm[i][0],
exclude_wlm[i][1], color='r', alpha=0.1)
plt.show()
plt.close()
vprint(f"Weights for getting smooth spectrum:\n fit_median ={weight_fit_median}"
+ f"\n Fit_median_interpolated = {(1-weight_fit_median)}")
# (fit_median+fit_median_interpolated)/2 # Decide if fit_median or fit_median_interpolated
return weight_fit_median*fit_median + (1-weight_fit_median)*fit_median_interpolated
[docs]def vect_norm(a, b):
"""Compute the norm of two vectors."""
return np.sqrt(np.sum((a - b)**2, axis=-1))
[docs]def in_rectangle(pos, rectangle_pos):
"""Check if a point lies in the perimeter of a rectangle."""
in_rectangle = (
(pos[0] <= rectangle_pos[:, 0]).any()
& (pos[0] >= rectangle_pos[:, 0]).any()
& (pos[1] <= rectangle_pos[:, 1]).any()
& (pos[1] >= rectangle_pos[:, 1]).any()
)
return in_rectangle
[docs]def pixel_in_circle(pixel_pos, pixel_size, circle_pos, circle_radius):
"""Compute the area of a pixel within a circle.
Parameters
----------
- pixel_pos: tuple
Position of the lower left corner of the pixel
- pixel_size: float
Size of the pixel.
- circle_pos: tuple
Position of the circle centre
- circle_raidus: float
Radius of the circle.
Returns
-------
- area_pixel:
Area of the pixel contained within the circle
- area_fraction:
Fration of the circle area that overlaps with the pixel.
"""
pixel_vertices = (np.atleast_1d(pixel_pos)[np.newaxis, :]
+ pixel_size * np.array([[0, 0], [0, 1], [1, 0], [1, 1]])
)
# Find those pixels inside the circle
inside = vect_norm(pixel_vertices, np.array(circle_pos)[np.newaxis]
) < circle_radius
n_inside = np.count_nonzero(inside)
if n_inside == 0:
return 0, 0
pixel_center = np.array([pixel_pos[0] + pixel_size / 2,
pixel_pos[1] + pixel_size / 2])
pixel_area = pixel_size**2
vertex_in = pixel_vertices[inside].squeeze()
vertex_out = pixel_vertices[~inside].squeeze()
cr_s = circle_radius**2
circle_area = np.pi * cr_s
if n_inside == 1:
x_cross = (cr_s - (vertex_in[1] - circle_pos[1])**2)**0.5
x_cross = np.array([[circle_pos[0] + x_cross, vertex_in[1]],
[circle_pos[0] - x_cross, vertex_in[1]]])
xcross_pt = np.argmin(vect_norm(pixel_center[np.newaxis, :], x_cross))
x_cross = x_cross[xcross_pt]
y_cross = (cr_s - (vertex_in[0] - circle_pos[0])**2)**0.5
y_cross = np.array([[vertex_in[0], circle_pos[1] + y_cross],
[vertex_in[0], circle_pos[1] - y_cross]])
ycross_pt = np.argmin(vect_norm(pixel_center[np.newaxis, :], y_cross))
y_cross = y_cross[ycross_pt]
arc_length = vect_norm(x_cross, y_cross)
phi = np.arccos(
(2*circle_radius**2 - arc_length**2) / (2 * circle_radius**2))
# Area of the triangle formed by the circle center and the intersection
area_triangle = 0.5 * circle_radius**2 * np.sin(phi)
area_sector = phi / 2 * circle_radius**2
# Area of the triangle within the pixel surface inside the circle
area_pixel_triangle = 0.5 * np.sqrt(
(vertex_in[0] - x_cross[0])**2 * (vertex_in[1] - y_cross[1])**2)
area_pixel = area_sector - area_triangle + area_pixel_triangle
elif n_inside == 2:
parallel_axis = np.where(vertex_in[0] == vertex_in[1])[0][0]
perpendicular_axis = [0, 1]
perpendicular_axis.remove(parallel_axis)
dist = np.abs(vertex_in[0][perpendicular_axis
] - vertex_in[1][perpendicular_axis]).squeeze()
intersection_points = np.zeros((2, 2))
for i, v in enumerate(vertex_in):
x_cross = (cr_s - (v[1] - circle_pos[1])**2)**0.5
x_cross = np.array([[circle_pos[0] + x_cross, v[1]],
[circle_pos[0] - x_cross, v[1]]])
xcross_pt = np.argmin(
vect_norm(pixel_center[np.newaxis, :], x_cross))
x_cross = x_cross[xcross_pt]
if in_rectangle(x_cross, pixel_vertices):
intersection_points[i] = x_cross
continue
y_cross = (cr_s - (v[0] - circle_pos[0])**2)**0.5
y_cross = np.array([[v[0], circle_pos[1] + y_cross],
[v[0], circle_pos[1] - y_cross]])
ycross_pt = np.argmin(
vect_norm(pixel_center[np.newaxis, :], y_cross))
y_cross = y_cross[ycross_pt]
if in_rectangle(y_cross, pixel_vertices):
intersection_points[i] = y_cross
# Parallel sides of the trapezium
a = np.abs(
vertex_in[0][parallel_axis]
- intersection_points[0][parallel_axis])
b = np.abs(
vertex_in[1][parallel_axis]
- intersection_points[1][parallel_axis])
area_trapezoid = dist * (a + b) / 2
arc_length = vect_norm(intersection_points[0], intersection_points[1])
phi = np.arccos(
(2*circle_radius**2 - arc_length**2) / (2 * circle_radius**2))
area_triangle = 0.5 * cr_s * np.sin(phi)
area_sector = phi / 2 * cr_s
area_pixel = area_sector - area_triangle + area_trapezoid
elif n_inside == 3:
dist_to_center = vect_norm(vertex_in, np.array(circle_pos)[np.newaxis])
corner_pos = np.argmin(dist_to_center)
intersection_points = np.zeros((2, 2))
for i, v in enumerate(np.delete(vertex_in, corner_pos, axis=0)):
x_cross = (cr_s - (v[1] - circle_pos[1])**2)**0.5
x_cross = np.array([[circle_pos[0] + x_cross, v[1]],
[circle_pos[0] - x_cross, v[1]]])
xcross_pt = np.argmin(
vect_norm(pixel_center[np.newaxis, :], x_cross))
x_cross = x_cross[xcross_pt]
if in_rectangle(x_cross, pixel_vertices):
intersection_points[i] = x_cross
continue
y_cross = (cr_s - (v[0] - circle_pos[0])**2)**0.5
y_cross = np.array([[v[0], circle_pos[1] + y_cross],
[v[0], circle_pos[1] - y_cross]])
ycross_pt = np.argmin(
vect_norm(pixel_center[np.newaxis, :], y_cross))
y_cross = y_cross[ycross_pt]
if in_rectangle(y_cross, pixel_vertices):
intersection_points[i] = y_cross
arc_length = vect_norm(intersection_points[0], intersection_points[1])
phi = np.arccos(
(2*circle_radius**2 - arc_length**2) / (2 * circle_radius**2))
area_inner_triangle = 0.5 * cr_s * np.sin(phi)
area_sector = phi / 2 * cr_s
area_outer_triangle = 0.5 * (
vect_norm(intersection_points[0], vertex_out)
* vect_norm(intersection_points[1], vertex_out))
area_pixel = (pixel_area - area_outer_triangle
+ (area_sector - area_inner_triangle))
else:
area_pixel = pixel_size**2
area_fraction = area_pixel / (circle_area + 1e-100)
return area_pixel, area_fraction
# ----------------------------------------------------------------------------------------------------------------------
# Models and fitting
# ----------------------------------------------------------------------------------------------------------------------
[docs]def cumulative_1d_sky(r2, sky_brightness):
"""
1D cumulative sky brightness. F_sky = 4*pi*r2 * B_sky
Parameters
----------
r2 : np.array(float)
Square radius from origin.
sky_brightness : float
Sky surface brightness.
Returns
-------
cumulative_sky_brightness : np.array(float)
Cumulative sky brightness.
"""
return np.pi * r2 * sky_brightness
[docs]def cumulative_1d_moffat(r2, l_star, alpha2, beta):
"""
Cumulative Moffat ligth profile.
Parameters
----------
r2 : np.array(float)
Square radius with respect to the profile centre.
l_star : float
Total luminosity integrating from 0 to inf.
alpha2 : float
Characteristic square radius.
beta : float
Power-low slope
Returns
-------
cum_moffat_prof: np.array(float)
Cumulative Moffat profile
"""
return l_star * (1 - np.power(1 + (r2 / alpha2), -beta))
[docs]def cumulative_1d_moffat_sky(r2, l_star, alpha2, beta, sky_brightness):
"""Combined model of cumulative_1d_moffat and cumulative_1d_sky."""
return cumulative_1d_sky(r2, sky_brightness) + cumulative_1d_moffat(r2, l_star, alpha2, beta)
[docs]def fit_moffat(r2_growth_curve, f_growth_curve,
f_guess, r2_half_light, r_max, plot=False):
"""
Fits a Moffat profile to a flux growth curve
as a function of radius squared,
cutting at to r_max (in units of the half-light radius),
provided an initial guess of the total flux and half-light radius squared.
# TODO
Parameters
----------
r2_growth_curve : TYPE
DESCRIPTION.
F_growth_curve : TYPE
DESCRIPTION.
F_guess : TYPE
DESCRIPTION.
r2_half_light : TYPE
DESCRIPTION.
r_max : TYPE
DESCRIPTION.
plot : Boolean, optional
If True generates and shows the plots. The default is False.
Returns
-------
fit : TYPE
DESCRIPTION.
"""
index_cut = np.searchsorted(r2_growth_curve, r2_half_light * r_max ** 2)
fit, cov = optimize.curve_fit(cumulative_1d_moffat,
r2_growth_curve[:index_cut], f_growth_curve[:index_cut],
p0=(f_guess, r2_half_light, 1)
)
if plot:
r_norm = np.sqrt(np.array(r2_growth_curve) / r2_half_light)
plt.plot(r_norm, cumulative_1d_moffat(np.array(r2_growth_curve),
fit[0], fit[1], fit[2]) / fit[0], ':')
return fit
[docs]def gaussian_2d(xy, amplitude, x0, y0, sigma_x, sigma_y, offset):
x, y = xy
exponent = -0.5 * (((x - x0) / sigma_x) ** 2 + ((y - y0) / sigma_y) ** 2)
return amplitude * np.exp(exponent) + offset
# =============================================================================
# Lines
# =============================================================================
lines = {
# Balmer
'hepsilon': 3970.1,
'hdelta': 4101.7,
'hgamma': 4340.4,
'hbeta': 4861.3,
'halpha': 6562.79,
# K
'K': 3934.777,
'H': 3969.588,
'Mg': 5176.7,
'Na': 5895.6,
'CaII1': 8500.36,
'CaII2': 8544.44,
'CaII3': 8664.52,
}
[docs]def mask_lines(wave_array, width=30, lines=lines.values()):
mask = np.ones_like(wave_array, dtype=bool)
for line in lines:
mask[(wave_array < line + width) & (wave_array > line - width)] = False
return mask
# =============================================================================
# Mr Krtxo \(ﾟ▽ﾟ)/
# ... Paranoy@ Rulz! ;^D