def __init__(self, initial, image_list, mask_list=None, invert=False,
square=False, outdir="", centers=None, psf_hw=13, kernel=None,
psfreg=0., sceneL2=0.0, dc=0.0, light=False, hdu=0):
# Metadata.
self.image_list = image_list
if mask_list is not None:
self.mask_list = dict([(k, mask_list[i])
for i, k in enumerate(image_list)])
else:
self.mask_list = {}
self.invert = invert
self.square = square
self.outdir = os.path.abspath(outdir)
self.psf_hw = psf_hw
self.psfreg = psfreg
self.sceneL2 = sceneL2
self.dc = dc
self.light = False
self.hdu = hdu
# Sort out the center vector and save it as a dictionary associated
# with specific filenames.
self.centers = centers
if centers is not None:
self.centers = dict([(image_list[i], centers[i])
for i in range(len(image_list))])
# Inference parameters.
self.sky = 0
self.scene = np.array(initial)
# 'Sky'-subtract the initial scene.
self.scene -= np.median(self.scene)
# Deal with the masked pixels if there are any in the initial scene
# by setting them to the 'sky' level.
self.scene[np.isnan(self.scene)] = 0.0
# Check the dimensions of the initial scene and set the size.
shape = self.scene.shape
assert shape[0] == shape[1], "The initial scene needs to be square."
self.size = shape[0] - 2 * self.psf_hw
# The 'kernel' used for 'light deconvolution'.
if kernel is None:
self.kernel = np.exp(-0.5 * (np.arange(-5, 6)[:, None] ** 2
+ np.arange(-5, 6)[None, :] ** 2))
self.kernel /= np.sum(self.kernel)
else:
self.kernel = kernel
# Index gymnastics.
self.scene_mask = utils.unravel_scene(self.size + 2 * self.psf_hw,
self.psf_hw)
self.psf_rows, self.psf_cols = \
utils.unravel_psf(self.size + 2 * self.psf_hw, self.psf_hw)
def test_unravel_psf(self):
"""
Test to make sure that the PSF unraveling yields the correct
results.
"""
S, P = 4, 1
rows, cols = utils.unravel_psf(S, P)
# Calculate the brute force unraveled scene.
b_cols = np.array(
[
0,
1,
2,
4,
5,
6,
8,
9,
10,
1,
2,
3,
5,
6,
7,
9,
10,
11,
4,
5,
6,
8,
9,
10,
12,
13,
14,
5,
6,
7,
9,
10,
11,
13,
14,
15,
],
dtype=int,
)
b_rows = np.concatenate([k * np.ones((2 * P + 1) ** 2, dtype=int) for k in range(4)])
assert np.all(rows == b_rows) and np.all(cols == b_cols)
def test_unravel_psf(self):
"""
Test to make sure that the PSF unraveling yields the correct
results.
"""
S, P = 4, 1
rows, cols = utils.unravel_psf(S, P)
# Calculate the brute force unraveled scene.
b_cols = np.array([
0, 1, 2, 4, 5, 6, 8, 9, 10, 1, 2, 3, 5, 6, 7, 9, 10, 11, 4, 5, 6,
8, 9, 10, 12, 13, 14, 5, 6, 7, 9, 10, 11, 13, 14, 15
],
dtype=int)
b_rows = np.concatenate(
[k * np.ones((2 * P + 1)**2, dtype=int) for k in range(4)])
assert np.all(rows == b_rows) and np.all(cols == b_cols)
