def filter_median(x, length=10, mask=None, partition=None, out=None):
"""
Median filtering, O(1) in window length
"""
from pyoperators.utils import strelapsed
time0 = time.time()
if out is None:
out = x.copy()
if mask is not None :
mask = np.ascontiguousarray(mask, np.bool8).view(np.int8)
elif hasattr(x, 'mask') and x.mask is not None and \
x.mask is not np.ma.nomask:
mask = x.mask.view(np.int8)
else:
mask = np.zeros(x.shape, np.int8)
n = x.shape[-1]
if partition is None:
partition = (n,)
status = tmf.filter_median(out.reshape((-1,n)).T,
mask.reshape((-1,n)).T, length,
np.array(partition, np.int32, ndmin=1))
if status != 0:
raise RuntimeError()
print(strelapsed(time0, 'Median filtering') + ' (length={0})'.format(length))
return out
def filter_median(x, length=10, mask=None, partition=None, out=None):
"""
Median filtering, O(1) in window length
"""
from pyoperators.utils import strelapsed
time0 = time.time()
if out is None:
out = x.copy()
if mask is not None:
mask = np.ascontiguousarray(mask, np.bool8).view(np.int8)
elif hasattr(x, 'mask') and x.mask is not None and \
x.mask is not np.ma.nomask:
mask = x.mask.view(np.int8)
else:
mask = np.zeros(x.shape, np.int8)
n = x.shape[-1]
if partition is None:
partition = (n, )
status = tmf.filter_median(
out.reshape((-1, n)).T,
mask.reshape((-1, n)).T, length, np.array(partition, np.int32,
ndmin=1))
if status != 0:
raise RuntimeError()
print(
strelapsed(time0, 'Median filtering') + ' (length={0})'.format(length))
return out
def _deglitch(tod, projection, nsigma, method):
time0 = time.time()
if method not in ('std', 'mad'):
raise ValueError('Invalid deglitching method.')
func = {'std': tmf.deglitch_l2b_std, 'mad': tmf.deglitch_l2b_mad}[method]
if isinstance(projection, ProjectionBaseOperator):
matrix = projection.matrix
npixels_per_sample = matrix.shape[-1]
elif hasattr(projection, 'operands') and \
all([isinstance(p, ProjectionBaseOperator) for p in projection.operands]):
ops = projection.operands
npixels_per_sample = max(p.matrix.shape[-1] for p in ops)
if tod.shape[-1] != sum(p.matrix.shape[-2] for p in ops):
raise ValueError('The tod has an incompatible shape.')
matrix = np.zeros(tod.shape + (npixels_per_sample, ),
ops[0].matrix.dtype)
matrix['index'] = -1
dest = 0
for p in ops:
nsamples = p.matrix.shape[-2]
matrix[..., dest:dest + nsamples, :p.matrix.shape[-1]] = p.matrix
dest += nsamples
else:
raise TypeError('The operator is not a projection.')
ny, nx = projection.shapein
if tod.mask is None:
mask = np.zeros(tod.shape, np.bool8)
else:
mask = tod.mask.copy()
percent = func(
matrix.view(int).ravel(), nx, ny, tod.T,
mask.view(np.int8).T, nsigma, npixels_per_sample)
# IFORT stores .true. as -1, it conflicts with numpy algebra on booleans
np.abs(mask, mask)
print(
strelapsed(time0, 'Deglitching ({0})'.format(method)) +
' (flagged samples: {0}%)'.format(percent))
return mask
def _deglitch(tod, projection, nsigma, method):
time0 = time.time()
if method not in ('std', 'mad'):
raise ValueError('Invalid deglitching method.')
func = {'std':tmf.deglitch_l2b_std, 'mad':tmf.deglitch_l2b_mad}[method]
if isinstance(projection, ProjectionBaseOperator):
matrix = projection.matrix
npixels_per_sample = matrix.shape[-1]
elif hasattr(projection, 'operands') and \
all([isinstance(p, ProjectionBaseOperator) for p in projection.operands]):
ops = projection.operands
npixels_per_sample = max(p.matrix.shape[-1] for p in ops)
if tod.shape[-1] != sum(p.matrix.shape[-2] for p in ops):
raise ValueError('The tod has an incompatible shape.')
matrix = np.zeros(tod.shape+(npixels_per_sample,), ops[0].matrix.dtype)
matrix['index'] = -1
dest = 0
for p in ops:
nsamples = p.matrix.shape[-2]
matrix[...,dest:dest+nsamples,:p.matrix.shape[-1]] = p.matrix
dest += nsamples
else:
raise TypeError('The operator is not a projection.')
ny, nx = projection.shapein
if tod.mask is None:
mask = np.zeros(tod.shape, np.bool8)
else:
mask = tod.mask.copy()
percent = func(matrix.view(int).ravel(), nx, ny, tod.T,
mask.view(np.int8).T, nsigma, npixels_per_sample)
# IFORT stores .true. as -1, it conflicts with numpy algebra on booleans
np.abs(mask, mask)
print(strelapsed(time0, 'Deglitching ({0})'.format(method)) +
' (flagged samples: {0}%)'.format(percent))
return mask
