def test_reshape(aia171_test_map, shape):
imagebytwo = reshape_image_to_4d_superpixel(aia171_test_map.data, (2, 2))
assert imagebytwo.shape == (shape[0] / 2, 2, shape[1] / 2, 2)
with pytest.raises(ValueError) as error_msg:
reshape_image_to_4d_superpixel(aia171_test_map.data, (3, 3))
assert 'New dimensions must divide original image size exactly.' in str(
error_msg.value)
def test_reshape(aia171_test_map, shape):
for factor in np.arange(2, 10):
if np.all((shape % factor) == 0):
# it is a factor therefore should work
reshape_image_to_4d_superpixel(aia171_test_map.data, shape/factor)
else:
# it is not a factor so function should raise an error
with pytest.raises(ValueError):
reshape_image_to_4d_superpixel(aia171_test_map.data, shape/factor)
def test_reshape(aia171_test_map, shape):
for factor in np.arange(2, 10):
if np.all((shape % factor) == 0):
# it is a factor therefore should work
reshape_image_to_4d_superpixel(aia171_test_map.data,
shape / factor)
else:
# it is not a factor so function should raise an error
with pytest.raises(ValueError):
reshape_image_to_4d_superpixel(aia171_test_map.data,
shape / factor)
def superpixel(self, dimensions, method="sum"):
"""Returns a new map consisting of superpixels formed from the
original data. Useful for increasing signal to noise ratio in images.
Parameters
----------
dimensions : tuple
One superpixel in the new map is equal to (dimension[0],
dimension[1]) pixels of the original map
Note: the first argument corresponds to the 'x' axis and the second
argument corresponds to the 'y' axis.
method : {'sum' | 'average'}
What each superpixel represents compared to the original data
* sum - add up the original data
* average - average the sum over the number of original pixels
Returns
-------
out : Map
A new Map which has superpixels of the required size.
References
----------
| http://mail.scipy.org/pipermail/numpy-discussion/2010-July/051760.html
"""
# Note: because the underlying ndarray is transposed in sense when
# compared to the Map, the ndarray is transposed, resampled, then
# transposed back
# Note: "center" defaults to True in this function because data
# coordinates in a Map are at pixel centers
# Make a copy of the original data and perform reshaping
reshaped = reshape_image_to_4d_superpixel(self.data.copy().T, dimensions)
if method == "sum":
new_data = reshaped.sum(axis=3).sum(axis=1)
elif method == "average":
new_data = (reshaped.sum(axis=3).sum(axis=1)) / np.float32(dimensions[0] * dimensions[1])
new_data = new_data.T
# Update image scale and number of pixels
new_map = deepcopy(self)
new_meta = new_map.meta
# Note that 'x' and 'y' correspond to 1 and 0 in self.shape,
# respectively
new_nx = self.shape[1] / dimensions[0]
new_ny = self.shape[0] / dimensions[1]
# Update metadata
new_meta["cdelt1"] = dimensions[0] * self.scale["x"]
new_meta["cdelt2"] = dimensions[1] * self.scale["y"]
new_meta["crpix1"] = (new_nx + 1) / 2.0
new_meta["crpix2"] = (new_ny + 1) / 2.0
new_meta["crval1"] = self.center["x"]
new_meta["crval2"] = self.center["y"]
# Create new map instance
new_map.data = new_data
return new_map
def test_reshape(aia171_test_map, shape):
def _n(a, b, c):
return int(np.floor((a-b)/c))
# Dimension divides the array shape exactly with no remainder
im = reshape_image_to_4d_superpixel(aia171_test_map.data, (2, 2), (0, 0))
assert im.shape == (shape[0]/2, 2, shape[1]/2, 2)
# Dimension divides the array shape exactly with remainder
im = reshape_image_to_4d_superpixel(aia171_test_map.data, (7, 5), (0, 0))
assert im.shape == (np.int(shape[0]/7), 7, np.int(shape[1]/5), 5)
# Dimension divides the array shape exactly with no remainder, and there is
# an offset
im = reshape_image_to_4d_superpixel(aia171_test_map.data, (2, 2), (1, 1))
assert im.shape == (np.int(shape[0]/2) - 1, 2, np.int(shape[1]/2) - 1, 2)
# Dimension divides the array shape exactly with remainder, and there is
# an offset
d = (9, 7)
o = (1, 4)
im = reshape_image_to_4d_superpixel(aia171_test_map.data, d, o)
assert im.shape == (_n(shape[0], o[0], d[0]), d[0],
_n(shape[1], o[1], d[1]), d[1])
im = reshape_image_to_4d_superpixel(aia171_test_map.data, d, o)
assert im.shape == (_n(shape[0], o[0], d[0]), d[0],
_n(shape[1], o[1], d[1]), d[1])
d = (9, 7)
o = (5, 4)
im = reshape_image_to_4d_superpixel(aia171_test_map.data, d, o)
assert im.shape == (_n(shape[0], o[0], d[0]), d[0],
_n(shape[1], o[1], d[1]), d[1])
d = (9, 7)
o = (4, 4)
im = reshape_image_to_4d_superpixel(aia171_test_map.data, d, o)
assert im.shape == (_n(shape[0], o[0], d[0]), d[0],
_n(shape[1], o[1], d[1]), d[1])
def test_reshape(aia171_test_map, shape):
def _n(a, b, c):
return int(np.floor((a - b) / c))
# Dimension divides the array shape exactly with no remainder
im = reshape_image_to_4d_superpixel(aia171_test_map.data, (2, 2), (0, 0))
assert im.shape == (shape[0] / 2, 2, shape[1] / 2, 2)
# Dimension divides the array shape exactly with remainder
im = reshape_image_to_4d_superpixel(aia171_test_map.data, (7, 5), (0, 0))
assert im.shape == (np.int(shape[0] / 7), 7, np.int(shape[1] / 5), 5)
# Dimension divides the array shape exactly with no remainder, and there is
# an offset
im = reshape_image_to_4d_superpixel(aia171_test_map.data, (2, 2), (1, 1))
assert im.shape == (np.int(shape[0] / 2) - 1, 2, np.int(shape[1] / 2) - 1,
2)
# Dimension divides the array shape exactly with remainder, and there is
# an offset
d = (9, 7)
o = (1, 4)
im = reshape_image_to_4d_superpixel(aia171_test_map.data, d, o)
assert im.shape == (_n(shape[0], o[0],
d[0]), d[0], _n(shape[1], o[1], d[1]), d[1])
im = reshape_image_to_4d_superpixel(aia171_test_map.data, d, o)
assert im.shape == (_n(shape[0], o[0],
d[0]), d[0], _n(shape[1], o[1], d[1]), d[1])
d = (9, 7)
o = (5, 4)
im = reshape_image_to_4d_superpixel(aia171_test_map.data, d, o)
assert im.shape == (_n(shape[0], o[0],
d[0]), d[0], _n(shape[1], o[1], d[1]), d[1])
d = (9, 7)
o = (4, 4)
im = reshape_image_to_4d_superpixel(aia171_test_map.data, d, o)
assert im.shape == (_n(shape[0], o[0],
d[0]), d[0], _n(shape[1], o[1], d[1]), d[1])
def superpixel(self, dimensions, method='sum'):
"""Returns a new map consisting of superpixels formed from the
original data. Useful for increasing signal to noise ratio in images.
Parameters
----------
dimensions : tuple
One superpixel in the new map is equal to (dimension[0],
dimension[1]) pixels of the original map
Note: the first argument corresponds to the 'x' axis and the second
argument corresponds to the 'y' axis.
method : {'sum' | 'average'}
What each superpixel represents compared to the original data
* sum - add up the original data
* average - average the sum over the number of original pixels
Returns
-------
out : Map
A new Map which has superpixels of the required size.
References
----------
| http://mail.scipy.org/pipermail/numpy-discussion/2010-July/051760.html
"""
# Note: because the underlying ndarray is transposed in sense when
# compared to the Map, the ndarray is transposed, resampled, then
# transposed back
# Note: "center" defaults to True in this function because data
# coordinates in a Map are at pixel centers
# Make a copy of the original data and perform reshaping
reshaped = reshape_image_to_4d_superpixel(self.data.copy().T,
dimensions)
if method == 'sum':
new_data = reshaped.sum(axis=3).sum(axis=1)
elif method == 'average':
new_data = ((reshaped.sum(axis=3).sum(axis=1)) /
np.float32(dimensions[0] * dimensions[1]))
new_data = new_data.T
# Update image scale and number of pixels
new_map = deepcopy(self)
new_meta = new_map.meta
# Note that 'x' and 'y' correspond to 1 and 0 in self.shape,
# respectively
new_nx = self.shape[1] / dimensions[0]
new_ny = self.shape[0] / dimensions[1]
# Update metadata
new_meta['cdelt1'] = dimensions[0] * self.scale['x']
new_meta['cdelt2'] = dimensions[1] * self.scale['y']
new_meta['crpix1'] = (new_nx + 1) / 2.
new_meta['crpix2'] = (new_ny + 1) / 2.
new_meta['crval1'] = self.center['x']
new_meta['crval2'] = self.center['y']
# Create new map instance
new_map.data = new_data
return new_map
def superpixel(self, dimensions, method='sum'):
"""Returns a new map consisting of superpixels formed from the
original data. Useful for increasing signal to noise ratio in images.
Parameters
----------
dimensions : tuple
One superpixel in the new map is equal to (dimension[0],
dimension[1]) pixels of the original map
Note: the first argument corresponds to the 'x' axis and the second
argument corresponds to the 'y' axis.
method : {'sum' | 'average'}
What each superpixel represents compared to the original data
* sum - add up the original data
* average - average the sum over the number of original pixels
Returns
-------
out : Map
A new Map which has superpixels of the required size.
References
----------
| http://mail.scipy.org/pipermail/numpy-discussion/2010-July/051760.html
"""
# Note: because the underlying ndarray is transposed in sense when
# compared to the Map, the ndarray is transposed, resampled, then
# transposed back
# Note: "center" defaults to True in this function because data
# coordinates in a Map are at pixel centers
# Make a copy of the original data and perform reshaping
reshaped = reshape_image_to_4d_superpixel(self.data.copy(),
[dimensions.value[1], dimensions.value[0]])
if method == 'sum':
new_data = reshaped.sum(axis=3).sum(axis=1)
elif method == 'average':
new_data = ((reshaped.sum(axis=3).sum(axis=1)) /
np.float32(dimensions[0] * dimensions[1]))
# Update image scale and number of pixels
new_map = deepcopy(self)
new_meta = new_map.meta
new_nx = (self.dimensions[0] / dimensions[0]).value
new_ny = (self.dimensions[1] / dimensions[1]).value
# Update metadata
new_meta['cdelt1'] = (dimensions[0] * self.scale.x).value
new_meta['cdelt2'] = (dimensions[1] * self.scale.y).value
if 'CD1_1' in new_meta:
new_meta['CD1_1'] *= dimensions[0].value
new_meta['CD2_1'] *= dimensions[0].value
new_meta['CD1_2'] *= dimensions[1].value
new_meta['CD2_2'] *= dimensions[1].value
new_meta['crpix1'] = (new_nx + 1) / 2.
new_meta['crpix2'] = (new_ny + 1) / 2.
new_meta['crval1'] = self.center.x.value
new_meta['crval2'] = self.center.y.value
# Create new map instance
new_map.data = new_data
return new_map
def test_reshape():
assert reshape_image_to_4d_superpixel(AIA_MAP.data, (512, 512)) != None
assert reshape_image_to_4d_superpixel(AIA_MAP.data, (600, 512)) == None
assert reshape_image_to_4d_superpixel(AIA_MAP.data, (512, 600)) == None
def superpixel(self, dimensions, method='sum'):
"""Returns a new map consisting of superpixels formed from the
original data. Useful for increasing signal to noise ratio in images.
Parameters
----------
dimensions : tuple
One superpixel in the new map is equal to (dimension[0],
dimension[1]) pixels of the original map
Note: the first argument corresponds to the 'x' axis and the second
argument corresponds to the 'y' axis.
method : {'sum' | 'average'}
What each superpixel represents compared to the original data
* sum - add up the original data
* average - average the sum over the number of original pixels
Returns
-------
out : Map
A new Map which has superpixels of the required size.
References
----------
| http://mail.scipy.org/pipermail/numpy-discussion/2010-July/051760.html
"""
from sunpy.image.rescale import reshape_image_to_4d_superpixel
# Note: because the underlying ndarray is transposed in sense when
# compared to the Map, the ndarray is transposed, resampled, then
# transposed back
# Note: "center" defaults to True in this function because data
# coordinates in a Map are at pixel centers
# Make a copy of the original data and perform reshaping
reshaped = reshape_image_to_4d_superpixel(np.asarray(self).copy().T,
dimensions)
if method == 'sum':
data = reshaped.sum(axis=3).sum(axis=1)
elif method == 'average':
data = ((reshaped.sum(axis=3).sum(axis=1)) /
np.float32(dimensions[0] * dimensions[1]))
#data = resample(np.asarray(self).copy().T, dimensions,
# method, center=True)
# Update image scale and number of pixels
header = self._original_header.copy()
# Note that 'x' and 'y' correspond to 1 and 0 in self.shape,
# respectively
new_nx = self.shape[1] / dimensions[0]
new_ny = self.shape[0] / dimensions[1]
# Create new map instance
new_map = self.__class__(data.T, header)
# Update metadata
new_map.scale['x'] = dimensions[0] * self.scale['x']
new_map.scale['y'] = dimensions[1] * self.scale['y']
new_map.reference_pixel['x'] = (new_nx + 1) / 2.
new_map.reference_pixel['y'] = (new_ny + 1) / 2.
new_map.reference_coordinate['x'] = self.center['x']
new_map.reference_coordinate['y'] = self.center['y']
return new_map
def test_reshape(): assert reshape_image_to_4d_superpixel(AIA_MAP.data, (512, 512)) != None assert reshape_image_to_4d_superpixel(AIA_MAP.data, (600, 512)) == None assert reshape_image_to_4d_superpixel(AIA_MAP.data, (512, 600)) == None
def test_reshape(aia171_test_map, shape):
imagebytwo = reshape_image_to_4d_superpixel(aia171_test_map.data, (2, 2))
assert imagebytwo.shape == (shape[0]/2, 2, shape[1]/2, 2)
with pytest.raises(ValueError) as error_msg:
reshape_image_to_4d_superpixel(aia171_test_map.data, (3, 3))
assert 'New dimensions must divide original image size exactly.' in str(error_msg.value)
