def background(self):
"""Compute the full background map."""
from dials.algorithms.image.filter import mean_filter
self._background /= self._count
self._gain *= self._mask.as_double()
return mean_filter(self._background, self._mask, self._kernel_size, 0)
def test_masked_mean_filter():
from scitbx.array_family import flex
from dials.algorithms.image.filter import mean_filter
# Create an image
image = flex.random_double(2000 * 2000)
image.reshape(flex.grid(2000, 2000))
mask = flex.random_bool(2000 * 2000, 0.99).as_int()
mask.reshape(flex.grid(2000, 2000))
# Calculate the summed area table
mask2 = mask.deep_copy()
mean = mean_filter(image, mask2, (3, 3), 1)
# For a selection of random points, ensure that the value is the
# sum of the area under the kernel
eps = 1e-7
for i in range(10000):
i = random.randint(10, 1990)
j = random.randint(10, 1990)
m1 = mean[j, i]
p = image[j - 3:j + 4, i - 3:i + 4]
m = mask[j - 3:j + 4, i - 3:i + 4]
if mask[j, i] == 0:
m2 = 0.0
else:
p = flex.select(p, flags=m)
mv = flex.mean_and_variance(flex.double(p))
m2 = mv.mean()
assert m1 == pytest.approx(m2, abs=eps)
def tst_mean_filter(self):
from dials.algorithms.image.filter import mean_filter
from scitbx.array_family import flex
from random import randint
# Create an image
image = flex.random_double(2000 * 2000)
image.reshape(flex.grid(2000, 2000))
# Calculate the summed area table
mean = mean_filter(image, (3, 3))
# For a selection of random points, ensure that the value is the
# sum of the area under the kernel
eps = 1e-7
for i in range(10000):
i = randint(10, 1990)
j = randint(10, 1990)
m1 = mean[j, i]
p = image[j - 3:j + 4, i - 3:i + 4]
mv = flex.mean_and_variance(p.as_1d())
m2 = mv.mean()
assert (abs(m1 - m2) <= eps)
# Test passed
print 'OK'
def tst_mean_filter(self):
from dials.algorithms.image.filter import mean_filter
from scitbx.array_family import flex
from random import randint
# Create an image
image = flex.random_double(2000 * 2000)
image.reshape(flex.grid(2000, 2000))
# Calculate the summed area table
mean = mean_filter(image, (3, 3))
# For a selection of random points, ensure that the value is the
# sum of the area under the kernel
eps = 1e-7
for i in range(10000):
i = randint(10, 1990)
j = randint(10, 1990)
m1 = mean[j,i]
p = image[j-3:j+4,i-3:i+4]
mv = flex.mean_and_variance(p.as_1d())
m2 = mv.mean()
assert(abs(m1 - m2) <= eps)
# Test passed
print 'OK'
def background(self):
'''Compute the full background map.'''
from dials.algorithms.image.filter import mean_filter
self._background /= self._count
self._gain *= self._mask.as_double()
return mean_filter(self._background, self._mask,
self._kernel_size, 0)
def gain(self):
"""Compute the full gain map."""
from dials.algorithms.image.filter import mean_filter
# Divide all gain values by count and smooth the gain
self._gain /= self._count
self._gain *= self._mask.as_double()
return mean_filter(self._gain, self._mask, self._kernel_size, 0)
def gain(self):
'''Compute the full gain map.'''
from dials.algorithms.image.filter import mean_filter
# Divide all gain values by count and smooth the gain
self._gain /= self._count
self._gain *= self._mask.as_double()
return mean_filter(self._gain, self._mask, self._kernel_size, 0)
def tst_masked_mean_filter(self):
from dials.algorithms.image.filter import mean_filter
from scitbx.array_family import flex
from random import randint
# Create an image
image = flex.random_double(2000 * 2000)
image.reshape(flex.grid(2000, 2000))
mask = flex.random_bool(2000 * 2000, 0.99).as_int()
mask.reshape(flex.grid(2000, 2000))
# Calculate the summed area table
mask2 = mask.deep_copy()
mean = mean_filter(image, mask2, (3, 3), 1)
# For a selection of random points, ensure that the value is the
# sum of the area under the kernel
eps = 1e-7
for i in range(10000):
i = randint(10, 1990)
j = randint(10, 1990)
m1 = mean[j,i]
p = image[j-3:j+4,i-3:i+4]
m = mask[j-3:j+4,i-3:i+4]
if mask[j,i] == 0:
m2 = 0.0
else:
p = flex.select(p, flags=m)
mv = flex.mean_and_variance(flex.double(p))
m2 = mv.mean()
s1 = flex.sum(flex.double(p))
s2 = flex.sum(m.as_1d())
assert(abs(m1 - m2) <= eps)
# Test passed
print 'OK'
def finalize(self, data, mask):
'''
Finalize the model
:param data: The data array
:param mask: The mask array
'''
from dials.algorithms.image.filter import median_filter, mean_filter
from dials.algorithms.image.fill_holes import diffusion_fill
from dials.algorithms.image.fill_holes import simple_fill
from dials.array_family import flex
# Print some image properties
sub_data = data.as_1d().select(mask.as_1d())
logger.info('Raw image statistics:')
logger.info(' min: %d' % int(flex.min(sub_data)))
logger.info(' max: %d' % int(flex.max(sub_data)))
logger.info(' mean: %d' % int(flex.mean(sub_data)))
logger.info('')
# Transform to polar
logger.info('Transforming image data to polar grid')
result = self.transform.to_polar(data, mask)
data = result.data()
mask = result.mask()
sub_data = data.as_1d().select(mask.as_1d())
logger.info('Polar image statistics:')
logger.info(' min: %d' % int(flex.min(sub_data)))
logger.info(' max: %d' % int(flex.max(sub_data)))
logger.info(' mean: %d' % int(flex.mean(sub_data)))
logger.info('')
# Filter the image to remove noise
if self.kernel_size > 0:
if self.filter_type == 'median':
logger.info('Applying median filter')
data = median_filter(data, mask, (self.kernel_size, 0))
sub_data = data.as_1d().select(mask.as_1d())
logger.info('Median polar image statistics:')
logger.info(' min: %d' % int(flex.min(sub_data)))
logger.info(' max: %d' % int(flex.max(sub_data)))
logger.info(' mean: %d' % int(flex.mean(sub_data)))
logger.info('')
elif self.filter_type == 'mean':
logger.info('Applying mean filter')
mask_as_int = mask.as_1d().as_int()
mask_as_int.reshape(mask.accessor())
data = mean_filter(data, mask_as_int, (self.kernel_size, 0), 1)
sub_data = data.as_1d().select(mask.as_1d())
logger.info('Mean polar image statistics:')
logger.info(' min: %d' % int(flex.min(sub_data)))
logger.info(' max: %d' % int(flex.max(sub_data)))
logger.info(' mean: %d' % int(flex.mean(sub_data)))
logger.info('')
else:
raise RuntimeError('Unknown filter_type: %s' % self.filter_type)
# Fill any remaining holes
logger.info("Filling holes")
data = simple_fill(data, mask)
data = diffusion_fill(data, mask, self.niter)
mask = flex.bool(data.accessor(), True)
sub_data = data.as_1d().select(mask.as_1d())
logger.info('Filled polar image statistics:')
logger.info(' min: %d' % int(flex.min(sub_data)))
logger.info(' max: %d' % int(flex.max(sub_data)))
logger.info(' mean: %d' % int(flex.mean(sub_data)))
logger.info('')
# Transform back
logger.info('Transforming image data from polar grid')
result = self.transform.from_polar(data, mask)
data = result.data()
mask = result.mask()
sub_data = data.as_1d().select(mask.as_1d())
logger.info('Final image statistics:')
logger.info(' min: %d' % int(flex.min(sub_data)))
logger.info(' max: %d' % int(flex.max(sub_data)))
logger.info(' mean: %d' % int(flex.mean(sub_data)))
logger.info('')
# Fill in any discontinuities
# FIXME NEED TO HANDLE DISCONTINUITY
# mask = ~self.transform.discontinuity()[:-1,:-1]
# data = diffusion_fill(data, mask, self.niter)
# Get and apply the mask
mask = self.experiment.imageset.get_mask(0)[0]
mask = mask.as_1d().as_int().as_double()
mask.reshape(data.accessor())
data *= mask
# Return the result
return data
def finalize(self, data, mask):
"""
Finalize the model
:param data: The data array
:param mask: The mask array
"""
from dials.algorithms.image.filter import median_filter, mean_filter
from dials.algorithms.image.fill_holes import diffusion_fill
from dials.algorithms.image.fill_holes import simple_fill
from dials.array_family import flex
# Print some image properties
sub_data = data.as_1d().select(mask.as_1d())
logger.info("Raw image statistics:")
logger.info(" min: %d" % int(flex.min(sub_data)))
logger.info(" max: %d" % int(flex.max(sub_data)))
logger.info(" mean: %d" % int(flex.mean(sub_data)))
logger.info("")
# Transform to polar
logger.info("Transforming image data to polar grid")
result = self.transform.to_polar(data, mask)
data = result.data()
mask = result.mask()
sub_data = data.as_1d().select(mask.as_1d())
logger.info("Polar image statistics:")
logger.info(" min: %d" % int(flex.min(sub_data)))
logger.info(" max: %d" % int(flex.max(sub_data)))
logger.info(" mean: %d" % int(flex.mean(sub_data)))
logger.info("")
# Filter the image to remove noise
if self.kernel_size > 0:
if self.filter_type == "median":
logger.info("Applying median filter")
data = median_filter(data,
mask, (self.kernel_size, 0),
periodic=True)
sub_data = data.as_1d().select(mask.as_1d())
logger.info("Median polar image statistics:")
logger.info(" min: %d" % int(flex.min(sub_data)))
logger.info(" max: %d" % int(flex.max(sub_data)))
logger.info(" mean: %d" % int(flex.mean(sub_data)))
logger.info("")
elif self.filter_type == "mean":
logger.info("Applying mean filter")
mask_as_int = mask.as_1d().as_int()
mask_as_int.reshape(mask.accessor())
data = mean_filter(data, mask_as_int, (self.kernel_size, 0), 1)
sub_data = data.as_1d().select(mask.as_1d())
logger.info("Mean polar image statistics:")
logger.info(" min: %d" % int(flex.min(sub_data)))
logger.info(" max: %d" % int(flex.max(sub_data)))
logger.info(" mean: %d" % int(flex.mean(sub_data)))
logger.info("")
else:
raise RuntimeError("Unknown filter_type: %s" %
self.filter_type)
# Fill any remaining holes
logger.info("Filling holes")
data = simple_fill(data, mask)
data = diffusion_fill(data, mask, self.niter)
mask = flex.bool(data.accessor(), True)
sub_data = data.as_1d().select(mask.as_1d())
logger.info("Filled polar image statistics:")
logger.info(" min: %d" % int(flex.min(sub_data)))
logger.info(" max: %d" % int(flex.max(sub_data)))
logger.info(" mean: %d" % int(flex.mean(sub_data)))
logger.info("")
# Transform back
logger.info("Transforming image data from polar grid")
result = self.transform.from_polar(data, mask)
data = result.data()
mask = result.mask()
sub_data = data.as_1d().select(mask.as_1d())
logger.info("Final image statistics:")
logger.info(" min: %d" % int(flex.min(sub_data)))
logger.info(" max: %d" % int(flex.max(sub_data)))
logger.info(" mean: %d" % int(flex.mean(sub_data)))
logger.info("")
# Fill in any discontinuities
mask = ~self.transform.discontinuity()[:-1, :-1]
data = diffusion_fill(data, mask, self.niter)
# Get and apply the mask
mask = self.experiment.imageset.get_mask(0)[0]
mask = mask.as_1d().as_int().as_double()
mask.reshape(data.accessor())
data *= mask
# Return the result
return data
