def test_type_exception():
"""
Test type exception.
"""
with pytest.raises(TypeError) as exc:
discretize_model(float(0), (-10, 11))
assert exc.value.args[0] == 'Model must be callable.'
def test_type_exception():
"""
Test type exception.
"""
with pytest.raises(TypeError) as exc:
discretize_model(float(0), (-10, 11))
assert exc.value.args[0] == 'Model must be callable.'
def test_discretize_oversample():
gauss_2D = Gaussian2D(amplitude=1.0,
x_mean=5.,
y_mean=125.,
x_stddev=0.75,
y_stddev=3)
values = discretize_model(gauss_2D,
x_range=[0, 10],
y_range=[100, 135],
mode='oversample',
factor=10)
vmax = np.max(values)
vmax_yx = np.unravel_index(values.argmax(), values.shape)
values_osf1 = discretize_model(gauss_2D,
x_range=[0, 10],
y_range=[100, 135],
mode='oversample',
factor=1)
values_center = discretize_model(gauss_2D,
x_range=[0, 10],
y_range=[100, 135],
mode='center')
assert values.shape == (35, 10)
assert_allclose(vmax, 0.927, atol=1e-3)
assert vmax_yx == (25, 5)
assert_allclose(values_center, values_osf1)
def test_float_y_range_exception():
def f(x, y):
return x ** 2 + y ** 2
with pytest.raises(ValueError) as exc:
discretize_model(f, (-10, 11), (-10.002, 11.23))
assert exc.value.args[0] == ("The difference between the upper an lower"
" limit of 'y_range' must be a whole number.")
def test_gaussian_sum_moments():
"""Check analytical against numerical solution.
"""
# We define three components with different flux, position and size
F_1, F_2, F_3 = 100, 200, 300
sigma_1, sigma_2, sigma_3 = 15, 10, 5
x_1, x_2, x_3 = 100, 120, 70
y_1, y_2, y_3 = 100, 90, 120
# Convert into non-normalized amplitude for astropy model
def A(F, sigma):
return F * 1 / (2 * np.pi * sigma ** 2)
# Define and evaluate models
f_1 = Gaussian2D(A(F_1, sigma_1), x_1, y_1, sigma_1, sigma_1)
f_2 = Gaussian2D(A(F_2, sigma_2), x_2, y_2, sigma_2, sigma_2)
f_3 = Gaussian2D(A(F_3, sigma_3), x_3, y_3, sigma_3, sigma_3)
F_1_image = discretize_model(f_1, (0, 200), (0, 200))
F_2_image = discretize_model(f_2, (0, 200), (0, 200))
F_3_image = discretize_model(f_3, (0, 200), (0, 200))
moments_num = measure_image_moments(F_1_image + F_2_image + F_3_image)
# Compute analytical values
cov_matrix = np.zeros((12, 12))
F = [F_1, F_2, F_3]
sigma = [sigma_1, sigma_2, sigma_3]
x = [x_1, x_2, x_3]
y = [y_1, y_2, y_3]
moments_ana, uncertainties = gaussian_sum_moments(F, sigma, x, y, cov_matrix)
assert_allclose(moments_ana, moments_num, 1e-6)
assert_allclose(uncertainties, 0)
def test_dim_exception_2d():
"""
Test dimension exception 2d.
"""
def f(x, y):
return x ** 2 + y ** 2
with pytest.raises(ValueError) as exc:
discretize_model(f, (-10, 11))
assert exc.value.args[0] == "y range not specified, but model is 2-d"
def test_dim_exception_1d():
"""
Test dimension exception 1d.
"""
def f(x):
return x ** 2
with pytest.raises(ValueError) as exc:
discretize_model(f, (-10, 11), (-10, 11))
assert exc.value.args[0] == "y range specified, but model is only 1-d."
def test_subpixel_gauss_2D():
"""
Test subpixel accuracy of the integrate mode with gaussian 2D model.
"""
gauss_2D = Gaussian2D(1, 0, 0, 0.1, 0.1)
values = discretize_model(gauss_2D, (-1, 2), (-1, 2), mode='integrate', factor=100)
assert_allclose(values.sum(), 2 * np.pi * 0.01, atol=0.00001)
def model_to_image(model, size, mode='center', factor=1, center=None):
"""
Converts 2D models into images using `astropy.convolution.utils.discretize_model`.
Parameters
----------
model : `~astropy.modeling.FittableModel` or callable.
Analytic model function to be discretized. Callables, which are not an
instances of `~astropy.modeling.FittableModel` are passed to
`~astropy.modeling.custom_model` and then evaluated.
size : int or tuple
The x and y size (in pixels) of the image in pixels (must be an whole number).
If only a single integer is provided, an image of equal x and y size
is generated. If tuple is provided (x_size, y_size) is assumed (N.B reverse of `numpy.array.shape` output).
mode : str, optional
One of the following modes (`astropy.convolution.utils.discretize_model`):
* ``'center'`` (default)
Discretize model by taking the value
at the center of the bin.
* ``'linear_interp'``
Discretize model by linearly interpolating
between the values at the corners of the bin.
For 2D models interpolation is bilinear.
* ``'oversample'``
Discretize model by taking the average
on an oversampled grid.
* ``'integrate'``
Discretize model by integrating the model
over the bin using `scipy.integrate.quad`.
Very slow.
factor : float or int
Factor of oversampling. Default = 1 (no oversampling).
center : tuple
(x, y) Coordinate of the center of the image (in pixels).
The origin of the image is defined as `origin = center - floor_divide(size, 2)`
(i.e the image will range from (origin -> origin + size)). If None, the origin
of the image is assumed to be at (0, 0) (i.e `center = floor_divide(size, 2)`).
Returns
-------
array : `numpy.array`
Model image
"""
x_size, y_size = _validate_image_size(size)
if center is None:
x_origin, y_origin = (0, 0)
else:
x_origin, y_origin = model_center_to_image_origin(center, size)
return discretize_model(model=model,
x_range=[x_origin, x_origin + x_size],
y_range=[y_origin, y_origin + y_size],
mode=mode,
factor=factor)
def test_subpixel_gauss_1D():
"""
Test subpixel accuracy of the integrate mode with gaussian 1D model.
"""
gauss_1D = Gaussian1D(1, 0, 0.1)
values = discretize_model(gauss_1D, (-1, 2), mode='integrate', factor=100)
assert_allclose(values.sum(), np.sqrt(2 * np.pi) * 0.1, atol=0.00001)
def test_discretize_callable_1d():
"""
Test discretize when a 1d function is passed.
"""
def f(x):
return x ** 2
y = discretize_model(f, (-5, 6))
assert_allclose(y, np.arange(-5, 6) ** 2)
def test_discretize_callable_2d():
"""
Test discretize when a 2d function is passed.
"""
def f(x, y):
return x ** 2 + y ** 2
actual = discretize_model(f, (-5, 6), (-5, 6))
y, x = (np.indices((11, 11)) - 5)
desired = x ** 2 + y ** 2
assert_allclose(actual, desired)
def test_gaussian_eval_1D(mode):
"""
Discretize Gaussian with different modes and check
if result is at least similar to Gaussian1D.eval().
"""
model = Gaussian1D(1, 0, 20)
x = np.arange(-100, 101)
values = model(x)
disc_values = discretize_model(model, (-100, 101), mode=mode)
assert_allclose(values, disc_values, atol=0.001)
def test_gaussian_eval_1D(mode):
"""
Discretize Gaussian with different modes and check
if result is at least similar to Gaussian1D.eval().
"""
model = Gaussian1D(1, 0, 20)
x = np.arange(-100, 101)
values = model(x)
disc_values = discretize_model(model, (-100, 101), mode=mode)
assert_allclose(values, disc_values, atol=0.001)
def test_pixel_sum_1D(model_class, mode):
"""
Test if the sum of all pixels corresponds nearly to the integral.
"""
if model_class == Box1D and mode == "center":
pytest.skip("Non integrating mode. Skip integral test.")
parameters = models_1D[model_class]
model = create_model(model_class, parameters)
values = discretize_model(model, models_1D[model_class]['x_lim'], mode=mode)
assert_allclose(values.sum(), models_1D[model_class]['integral'], atol=0.0001)
def test_gaussian_eval_2D(mode):
"""
Discretize Gaussian with different modes and check
if result is at least similar to Gaussian2D.eval()
"""
model = Gaussian2D(0.01, 0, 0, 1, 1)
x = np.arange(-2, 3)
y = np.arange(-2, 3)
x, y = np.meshgrid(x, y)
values = model(x, y)
disc_values = discretize_model(model, (-2, 3), (-2, 3), mode=mode)
assert_allclose(values, disc_values, atol=1e-2)
def test_gaussian_eval_2D(mode):
"""
Discretize Gaussian with different modes and check
if result is at least similar to Gaussian2D.eval()
"""
model = Gaussian2D(0.01, 0, 0, 1, 1)
x = np.arange(-2, 3)
y = np.arange(-2, 3)
x, y = np.meshgrid(x, y)
values = model(x, y)
disc_values = discretize_model(model, (-2, 3), (-2, 3), mode=mode)
assert_allclose(values, disc_values, atol=1e-2)
def test_gaussian_eval_2D_integrate_mode():
"""
Discretize Gaussian with integrate mode
"""
model_list = [Gaussian2D(.01, 0, 0, 2, 2),
Gaussian2D(.01, 0, 0, 1, 2),
Gaussian2D(.01, 0, 0, 2, 1)]
x = np.arange(-2, 3)
y = np.arange(-2, 3)
x, y = np.meshgrid(x, y)
for model in model_list:
values = model(x, y)
disc_values = discretize_model(model, (-2, 3), (-2, 3), mode='integrate')
assert_allclose(values, disc_values, atol=1e-2)
def overlap_image(request):
if request.param == 2:
close_tab = Table([[50., 53.], [50., 50.], [25., 25.]], names=['x_0', 'y_0', 'flux_0'])
elif request.param == 3:
close_tab = Table([[50., 55., 50.], [50., 50., 55.], [25., 25., 25.]],
names=['x_0', 'y_0', 'flux_0'])
else:
raise ValueError
# Add sources to test image
close_image = np.zeros((IMAGE_SIZE, IMAGE_SIZE))
for x, y, flux in close_tab:
close_model = Gaussian2D(flux / (2 * np.pi * GAUSSIAN_WIDTH ** 2),
x, y, GAUSSIAN_WIDTH, GAUSSIAN_WIDTH)
close_image += discretize_model(close_model, (0, IMAGE_SIZE), (0, IMAGE_SIZE),
mode='oversample')
return close_image
def evaluate_model(self, **kwargs):
"""Evaluate model by oversampling or taking the value at the center of the pixel.
"""
self._setup_model()
self.model_image = np.zeros_like(self.exposure, dtype=np.float64)
from astropy.convolution import utils
height, width = self.exposure.shape
for source_model in self.source_models:
source_model_image = utils.discretize_model(
source_model, (0, width), (0, height), **kwargs)
self.model_image += source_model_image
if self._compute_excess:
self.model_image = self.model_image * self.exposure
if self._apply_psf:
psf = self._create_psf(**kwargs)
from astropy.convolution import convolve
self.model_image = convolve(self.model_image, psf)
self.model_image *= self._flux_factor
def simulate_image(images):
'''pick image, pick source parameters, add gaussian source'''
n = np.random.choice(images.shape[2])
halfwidth = images.shape[0] // 2
theta = np.random.uniform() * 2. * np.pi
r = np.random.uniform() * halfwidth # pix
x_in = halfwidth + r * np.sin(theta) # pix
y_in = halfwidth + r * np.cos(theta) # pix
flux_in = 10 ** (np.random.uniform() * 7 + 1)
sigma_in = np.random.uniform() * 0.5 + 1.
val_in = {'id': [n], 'theta': [theta], 'r': [r], 'x': [x_in], 'y': [y_in],
'flux': [flux_in], 'sigma': [sigma_in]}
# Create test psf
psf_model = Gaussian2D(flux_in / (2 * np.pi * sigma_in ** 2), x_in,
y_in, sigma_in, sigma_in)
test_image = images[:, :, n] + discretize_model(psf_model,
(0, images.shape[0]),
(0, images.shape[1]),
mode='oversample')
return Table(val_in), test_image
def evaluate_model(self, **kwargs):
"""Evaluate model by oversampling or taking the value at the center of the pixel.
"""
self._setup_model()
self.model_image = np.zeros_like(self.exposure, dtype=np.float64)
from astropy.convolution import utils
height, width = self.exposure.shape
for source_model in self.source_models:
source_model_image = utils.discretize_model(source_model,
(0, width), (0, height), **kwargs)
self.model_image += source_model_image
if self._compute_excess:
self.model_image = self.model_image * self.exposure
if self._apply_psf:
psf = self._create_psf(**kwargs)
from astropy.convolution import convolve
self.model_image = convolve(self.model_image, psf)
self.model_image *= self._flux_factor
# tests previously written to psf_photometry
PSF_SIZE = 11
GAUSSIAN_WIDTH = 1.0
IMAGE_SIZE = 101
# Position and FLUXES of test sources
INTAB = Table(
[[50.0, 23, 12, 86], [50.0, 83, 80, 84], [np.pi * 10, 3.654, 20.0, 80 / np.sqrt(3)]], names=["x_0", "y_0", "flux_0"]
)
# Create test psf
psf_model = Gaussian2D(
1.0 / (2 * np.pi * GAUSSIAN_WIDTH ** 2), PSF_SIZE // 2, PSF_SIZE // 2, GAUSSIAN_WIDTH, GAUSSIAN_WIDTH
)
test_psf = discretize_model(psf_model, (0, PSF_SIZE), (0, PSF_SIZE), mode="oversample")
# Set up grid for test image
image = np.zeros((IMAGE_SIZE, IMAGE_SIZE))
# Add sources to test image
for x, y, flux in INTAB:
model = Gaussian2D(flux / (2 * np.pi * GAUSSIAN_WIDTH ** 2), x, y, GAUSSIAN_WIDTH, GAUSSIAN_WIDTH)
image += discretize_model(model, (0, IMAGE_SIZE), (0, IMAGE_SIZE), mode="oversample")
# Some tests require an image with wider sources.
WIDE_GAUSSIAN_WIDTH = 3.0
WIDE_INTAB = Table([[50, 23.2], [50.5, 1], [10, 20]], names=["x_0", "y_0", "flux_0"])
wide_image = np.zeros((IMAGE_SIZE, IMAGE_SIZE))
# Add sources to test image
HAS_SCIPY = False
PSF_SIZE = 11
GAUSSIAN_WIDTH = 1.
IMAGE_SIZE = 101
# Position and FLUXES of test sources
INTAB = Table([[50., 23, 12, 86], [50., 83, 80, 84],
[np.pi * 10, 3.654, 20., 80 / np.sqrt(3)]],
names=['x_0', 'y_0', 'flux_0'])
# Create test psf
psf_model = Gaussian2D(1. / (2 * np.pi * GAUSSIAN_WIDTH ** 2), PSF_SIZE // 2,
PSF_SIZE // 2, GAUSSIAN_WIDTH, GAUSSIAN_WIDTH)
test_psf = discretize_model(psf_model, (0, PSF_SIZE), (0, PSF_SIZE),
mode='oversample')
# Set up grid for test image
image = np.zeros((IMAGE_SIZE, IMAGE_SIZE))
# Add sources to test image
for x, y, flux in INTAB:
model = Gaussian2D(flux / (2 * np.pi * GAUSSIAN_WIDTH ** 2),
x, y, GAUSSIAN_WIDTH, GAUSSIAN_WIDTH)
image += discretize_model(model, (0, IMAGE_SIZE), (0, IMAGE_SIZE),
mode='oversample')
@pytest.mark.skipif('not HAS_SCIPY')
def test_subtract_psf():
"""Test subtract_psf."""
# tests previously written to psf_photometry
PSF_SIZE = 11
GAUSSIAN_WIDTH = 1.
IMAGE_SIZE = 101
# Position and FLUXES of test sources
INTAB = Table([[50., 23, 12, 86], [50., 83, 80, 84],
[np.pi * 10, 3.654, 20., 80 / np.sqrt(3)]],
names=['x_0', 'y_0', 'flux_0'])
# Create test psf
psf_model = Gaussian2D(1. / (2 * np.pi * GAUSSIAN_WIDTH**2), PSF_SIZE // 2,
PSF_SIZE // 2, GAUSSIAN_WIDTH, GAUSSIAN_WIDTH)
test_psf = discretize_model(psf_model, (0, PSF_SIZE), (0, PSF_SIZE),
mode='oversample')
# Set up grid for test image
image = np.zeros((IMAGE_SIZE, IMAGE_SIZE))
# Add sources to test image
for x, y, flux in INTAB:
model = Gaussian2D(flux / (2 * np.pi * GAUSSIAN_WIDTH**2), x, y,
GAUSSIAN_WIDTH, GAUSSIAN_WIDTH)
image += discretize_model(model, (0, IMAGE_SIZE), (0, IMAGE_SIZE),
mode='oversample')
# Some tests require an image with wider sources.
WIDE_GAUSSIAN_WIDTH = 3.
WIDE_INTAB = Table([[50, 23.2], [50.5, 1], [10, 20]],
names=['x_0', 'y_0', 'flux_0'])
except ImportError:
HAS_SCIPY = False
psf_size = 11
gaussian_width = 1.
image_size = 101.
# Position and fluxes of tes sources
positions = [(50, 50), (23, 83), (12, 80), (86, 84)]
fluxes = [np.pi * 10, 3.654, 20., 80 / np.sqrt(3)]
# Create test psf
psf_model = Gaussian2D(1. / (2 * np.pi * gaussian_width ** 2),
psf_size // 2, psf_size // 2, gaussian_width, gaussian_width)
test_psf = discretize_model(psf_model, (0, psf_size), (0, psf_size), mode='oversample')
# Set up grid for test image
image = np.zeros((image_size, image_size))
# Add sources to test image
for i, position in enumerate(positions):
x, y = position
model = Gaussian2D(fluxes[i] / (2 * np.pi * gaussian_width ** 2),
x, y, gaussian_width, gaussian_width)
image += discretize_model(model, (0, image_size), (0, image_size), mode='oversample')
def test_create_prf_mean():
"""
Check if create_prf works correctly on simulated data.
