def make_background_image():
"""Estimate background image.
See the `KernelBackgroundEstimator` tutorial and
documentation how it works, or the SciNeGHe 2014 proceeding
by Ellis Owen et al.
"""
radius = Angle(0.2, 'deg')
r_in = Angle(0.3, 'deg')
r_out = Angle(0.7, 'deg')
significance_threshold = 5
mask_dilation_radius = Angle(0.1, 'deg')
max_iterations = 3
hdu = fits.open(COUNTS_IMAGE)['COUNTS']
binsz = hdu.header['CDELT2']
images = KernelBackgroundEstimatorData(counts=hdu.data, header=hdu.header)
# TODO: we should have utility functions to initialise
# kernels with angles so that we don't have to convert to pix here.
source_kernel = binary_disk(radius=radius.deg/binsz)
background_kernel = binary_ring(r_in=r_in.deg/binsz,
r_out=r_out.deg/binsz)
estimator = KernelBackgroundEstimator(
images=images, source_kernel=source_kernel, background_kernel=background_kernel,
significance_threshold=significance_threshold,
mask_dilation_radius=mask_dilation_radius.deg/binsz,
)
print('Running background estimation ...')
estimator.run(max_iterations=max_iterations)
print('Writing {}'.format(MASK_IMAGE))
estimator.mask_image_hdu.writeto(MASK_IMAGE, clobber=True)
print('Writing {}'.format(BACKGROUND_IMAGE))
estimator.background_image_hdu.writeto(BACKGROUND_IMAGE, clobber=True)
from gammapy.detect import KernelBackgroundEstimator
# Parameters
CORRELATION_RADIUS = 10 # Pixels
SIGNIFICANCE_THRESHOLD = 5 # Sigma
MASK_DILATION_RADIUS = 0.5 * u.deg
# Load example images.
filename = ('$GAMMAPY_EXTRA/datasets/source_diffuse_separation/'
'galactic_simulations/fermi_counts.fits')
counts = SkyImage.read(filename)
center = SkyCoord(0, 0, frame='galactic', unit='deg')
images = SkyImageList()
images['counts'] = counts.cutout(center, (10 * u.deg, 80 * u.deg))
kernel_src = Tophat2DKernel(CORRELATION_RADIUS).array
kernel_bkg = np.ones((10, 150))
kbe = KernelBackgroundEstimator(
kernel_src=kernel_src,
kernel_bkg=kernel_bkg,
significance_threshold=SIGNIFICANCE_THRESHOLD,
mask_dilation_radius=MASK_DILATION_RADIUS,
)
result = kbe.run(images)
kbe.images_stack_show()
plt.show()
# Counts must be provided as an ImageHDU
counts = fits.open(filename)[0].data
header = fits.open(filename)[0].header
images = KernelBackgroundEstimatorData(counts=counts, header=header)
source_kernel = binary_disk(CORRELATION_RADIUS)
background_kernel = np.ones((10, 100))
# *** ITERATOR ***
kbe = KernelBackgroundEstimator(
images=images,
source_kernel=source_kernel,
background_kernel=background_kernel,
significance_threshold=SIGNIFICANCE_THRESHOLD,
mask_dilation_radius=MASK_DILATION_RADIUS,
)
n_iterations = 4
# *** RUN & PLOT ***
plt.figure(figsize=(8, 4))
for iteration in range(n_iterations):
kbe.run_iteration()
mask_hdu = kbe.mask_image_hdu
mask = mask_hdu.data[:, 1400:2000]
plt.subplot(n_iterations, 2, 2 * iteration + 1)
# Counts must be provided as an ImageHDU
counts = fits.open(filename)[0].data
header = fits.open(filename)[0].header
images = KernelBackgroundEstimatorData(counts=counts, header=header)
source_kernel = binary_disk(CORRELATION_RADIUS)
background_kernel = np.ones((10, 100))
# *** ITERATOR ***
kbe = KernelBackgroundEstimator(
images=images,
source_kernel=source_kernel,
background_kernel=background_kernel,
significance_threshold=SIGNIFICANCE_THRESHOLD,
mask_dilation_radius=MASK_DILATION_RADIUS,
)
n_iterations = 4
# *** RUN & PLOT ***
plt.figure(figsize=(8, 4))
for iteration in range(n_iterations):
kbe.run_iteration()
mask_hdu = kbe.mask_image_hdu
mask = mask_hdu.data[:, 1400:2000]
plt.subplot(n_iterations, 2, 2 * iteration + 1)