347 lines
12 KiB
Python
347 lines
12 KiB
Python
import numpy as np
|
|
from scipy.linalg import lu
|
|
|
|
import xarray as xr
|
|
|
|
class InvalidDimException(Exception):
|
|
"Raised when the program can not identify (index of images, x, y) axes."
|
|
def __init__(self, dims):
|
|
if len(dims)>3:
|
|
self.message = 'The input data must have two or three axes: (index of images(alternative), x, y)'
|
|
else:
|
|
self.message = 'Can not identify (index of images(alternative), x, y) from ' + str(dims)
|
|
super().__init__(self.message)
|
|
|
|
|
|
class DataSizeException(Exception):
|
|
"Raised when the shape of the data is not correct."
|
|
def __init__(self):
|
|
self.message = 'The input data size does not match.'
|
|
super().__init__(self.message)
|
|
|
|
|
|
class FringeRemoval():
|
|
"""
|
|
FRINGEREMOVAL - Fringe removal and noise reduction from absorption images.
|
|
Creates an optimal reference image for each absorption image in a set as
|
|
a linear combination of reference images, with coefficients chosen to
|
|
minimize the least-squares residuals between each absorption image and
|
|
the optimal reference image. The coefficients are obtained by solving a
|
|
linear set of equations using matrix inverse by LU decomposition.
|
|
|
|
Application of the algorithm is described in C. F. Ockeloen et al, Improved
|
|
detection of small atom numbers through image processing, arXiv:1007.2136 (2010).
|
|
|
|
Adapted from a MATLAB script copy provided by Guoxian Su.
|
|
Original Authors: Shannon Whitlock, Caspar Ockeloen
|
|
Reference: C. F. Ockeloen, A. F. Tauschinsky, R. J. C. Spreeuw, and
|
|
S. Whitlock, Improved detection of small atom numbers through
|
|
image processing, arXiv:1007.2136
|
|
May 2009;
|
|
|
|
"""
|
|
|
|
def __init__(self) -> None:
|
|
"""Initialize the class
|
|
"""
|
|
self.nimgsR = 0 # The number of the reference images
|
|
self.xdim = 0 # The shape of x axis
|
|
self.ydim = 0 # The shape of y axis
|
|
|
|
self._mask = None # The mask array to choose the region of interest for fringes removal
|
|
self._center = None # Set the mask array by center and span
|
|
self._span = None
|
|
|
|
self.reshape=True # If it is necessary to reshape the data from (index of images(alternative), x, y) to (y, x, index of images(alternative))
|
|
|
|
self.P = None
|
|
self.L = None
|
|
self.U = None
|
|
|
|
@property
|
|
def center(self):
|
|
"""The getter of the center of region of insterest (ROI)
|
|
|
|
:return: The center of region of insterest (ROI)
|
|
:rtype: tuple
|
|
"""
|
|
return self._center
|
|
|
|
@center.setter
|
|
def center(self, value):
|
|
"""The setter of the center of region of insterest (ROI)
|
|
|
|
:param value: The center of region of insterest (ROI)
|
|
:type value: tuple
|
|
"""
|
|
self._mask = None
|
|
self._center = value
|
|
|
|
@property
|
|
def span(self):
|
|
"""The getter of the span of region of insterest (ROI)
|
|
|
|
:return: The span of region of insterest (ROI)
|
|
:rtype: tuple
|
|
"""
|
|
return self._span
|
|
|
|
@span.setter
|
|
def span(self, value):
|
|
"""The setter of the span of region of insterest (ROI)
|
|
|
|
:param value: The span of region of insterest (ROI)
|
|
:type value: tuple
|
|
"""
|
|
self._mask = None
|
|
self._span = value
|
|
|
|
def reshape_data(self, data):
|
|
"""The function is to reshape the data to the correct shape.
|
|
In order to minimize the calculation time, the data has to have a shape of (y, x, index of images(alternative)).
|
|
However, usually the input data has a shape of (index of images(alternative), x, y).
|
|
It can also convert the xarray DataArray and Dataset to numpy array.
|
|
|
|
:param data: The input data.
|
|
:type data: xarray, numpy array or list
|
|
:raises InvalidDimException: Raised when the program can not identify (index of images, x, y) axes.
|
|
:raises InvalidDimException: Raised when the shape of the data is not correct.
|
|
:return: The data with correct shape
|
|
:rtype: xarray, numpy array or list
|
|
"""
|
|
|
|
if data is None:
|
|
return data
|
|
|
|
if isinstance(data, type(xr.DataArray())):
|
|
|
|
dims = data.dims
|
|
|
|
if len(dims)>3:
|
|
raise InvalidDimException(dims)
|
|
|
|
xAxis = None
|
|
yAxis = None
|
|
if len(dims) == 2:
|
|
imageAxis = ''
|
|
else:
|
|
imageAxis = None
|
|
|
|
for dim in dims:
|
|
if (dim == 'x') or ('_x' in dim):
|
|
xAxis = dim
|
|
elif (dim == 'y') or ('_y' in dim):
|
|
yAxis = dim
|
|
else:
|
|
imageAxis = dim
|
|
|
|
if (xAxis is None) or (yAxis is None) or (imageAxis is None):
|
|
raise InvalidDimException(dims)
|
|
|
|
if len(dims) == 2:
|
|
data = data.transpose(yAxis, xAxis)
|
|
else:
|
|
data = data.transpose(yAxis, xAxis, imageAxis)
|
|
|
|
data = data.to_numpy()
|
|
|
|
else:
|
|
data = np.array(data)
|
|
if len(data.shape) == 3:
|
|
data = np.swapaxes(data, 0, 2)
|
|
# data = np.swapaxes(data, 0, 1)
|
|
elif len(data.shape) == 2:
|
|
data = np.swapaxes(data, 0, 1)
|
|
|
|
return data
|
|
|
|
def _reshape_absorption_images(self, data):
|
|
|
|
if data is None:
|
|
return data
|
|
|
|
if isinstance(data, type(xr.DataArray())):
|
|
|
|
dims = data.dims
|
|
|
|
if len(dims)>3:
|
|
raise InvalidDimException(dims)
|
|
|
|
xAxis = None
|
|
yAxis = None
|
|
if len(dims) == 2:
|
|
imageAxis = ''
|
|
else:
|
|
imageAxis = None
|
|
|
|
for dim in dims:
|
|
if (dim == 'x') or ('_x' in dim):
|
|
xAxis = dim
|
|
elif (dim == 'y') or ('_y' in dim):
|
|
yAxis = dim
|
|
else:
|
|
imageAxis = dim
|
|
|
|
if (xAxis is None) or (yAxis is None) or (imageAxis is None):
|
|
raise InvalidDimException(dims)
|
|
|
|
if len(dims) == 2:
|
|
data = data.transpose(yAxis, xAxis)
|
|
else:
|
|
data = data.transpose(yAxis, xAxis, imageAxis)
|
|
|
|
self.nimgs = len(data[imageAxis])
|
|
|
|
data = data.stack(axis=[yAxis, xAxis])
|
|
|
|
else:
|
|
data = np.array(data)
|
|
if len(data.shape) == 3:
|
|
data = np.swapaxes(data, 0, 2)
|
|
# data = np.swapaxes(data, 0, 1)
|
|
elif len(data.shape) == 2:
|
|
data = np.swapaxes(data, 0, 1)
|
|
|
|
return data
|
|
|
|
@property
|
|
def referenceImages(self):
|
|
res = self._referenceImages.reshape(self.ydim, self.xdim, self.nimgsR)
|
|
res = np.swapaxes(res, 0, 2)
|
|
return res
|
|
|
|
@referenceImages.setter
|
|
def referenceImages(self, value):
|
|
if value is None:
|
|
self._referenceImages = None
|
|
return
|
|
|
|
if self.reshape:
|
|
value = self.reshape_data(value)
|
|
elif isinstance(value, type(xr.DataArray())):
|
|
value = value.to_numpy()
|
|
|
|
self.nimgsR = value.shape[2]
|
|
self.xdim = value.shape[1]
|
|
self.ydim = value.shape[0]
|
|
|
|
self._referenceImages = (value.reshape(self.xdim * self.ydim, self.nimgsR).astype(np.float32))
|
|
|
|
def add_reference_images(self, data):
|
|
"""Add a new reference images
|
|
|
|
:param data: The new reference image.
|
|
:type data: xarray, numpy array or list
|
|
:raises DataSizeException: Raised when the shape of the data is not correct.
|
|
"""
|
|
|
|
if self.reshape:
|
|
data = self.reshape_data(data)
|
|
elif isinstance(data, type(xr.DataArray())):
|
|
data = data.to_numpy()
|
|
|
|
if not ((data.shape[0]==self.ydim) and (data.shape[1]==self.xdim)):
|
|
raise DataSizeException
|
|
|
|
data = data.reshape(self.xdim * self.ydim)
|
|
|
|
self._referenceImages = np.append(self._referenceImages, data, axis=1)
|
|
|
|
def _remove_first_reference_images(self):
|
|
"""Remove the first reference images
|
|
"""
|
|
self._referenceImages = np.delete(self._referenceImages, 0, axis=1)
|
|
|
|
def update_reference_images(self, data):
|
|
"""Update the reference images set by removing the first one and adding a new one at the end.
|
|
|
|
:param data: The new reference image.
|
|
:type data: xarray, numpy array or list
|
|
"""
|
|
self._remove_first_reference_images()
|
|
self.add_reference_images(data)
|
|
|
|
self.decompose_referenceImages()
|
|
|
|
@property
|
|
def mask(self):
|
|
return self._mask
|
|
|
|
@mask.setter
|
|
def mask(self, value):
|
|
if self.reshape:
|
|
value = self.reshape_data(value)
|
|
elif isinstance(value, type(xr.DataArray())):
|
|
value = value.to_numpy()
|
|
|
|
if not ((value.shape[0]==self.ydim) and (value.shape[1]==self.xdim)):
|
|
raise DataSizeException
|
|
|
|
self._mask = value
|
|
|
|
self._center = None
|
|
self._span = None
|
|
|
|
self.k = np.where(self._mask.flatten() == 1)[0]
|
|
|
|
def _auto_mask(self):
|
|
mask = np.ones((self.ydim, self.xdim), dtype=np.uint8)
|
|
if not self._center is None:
|
|
x_start = int(self._center[0] - self._span[0] / 2)
|
|
x_end = int(self._center[0] + self._span[0] / 2)
|
|
y_end = int(self._center[1] + self._span[1] / 2)
|
|
y_start = int(self._center[1] - self._span[1] / 2)
|
|
|
|
mask[y_start:y_end, x_start:x_end] = 0
|
|
|
|
return mask
|
|
|
|
def decompose_referenceImages(self):
|
|
if self._mask is None:
|
|
self.mask = self._auto_mask()
|
|
self.P, self.L, self.U = lu(self._referenceImages[self.k, :].T @ self._referenceImages[self.k, :], permute_l = False, p_indices = True)
|
|
|
|
def _fringe_removal(self, absorptionImages):
|
|
|
|
b = self.temp @ absorptionImages[self.k]
|
|
c = np.linalg.solve(self.U, np.linalg.solve(self.L[self.P], b))
|
|
optrefimages = (self._referenceImages @ c)
|
|
|
|
return optrefimages
|
|
|
|
def fringe_removal(self, absorptionImages, referenceImages=None, mask=None, reshape=None, dask='parallelized'):
|
|
"""
|
|
This function will generate a 'fake' background images, which can help to remove the fringes.
|
|
|
|
Important: Please substract the drak images from the both of images with atoms and without atoms before using this function!!!
|
|
|
|
:param absorptionImages: A set of images with atoms in absorption imaging
|
|
:type absorptionImages: xarray, numpy array or list
|
|
:param referenceImages: A set of images without atoms in absorption imaging, defaults to None
|
|
:type referenceImages: xarray, numpy array or list, optional
|
|
:param mask: An array to choose the region of interest for fringes removal, defaults to None, defaults to None
|
|
:type mask: numpy array, optional
|
|
:param reshape: If it needs to reshape the data, defaults to None
|
|
:type reshape: bool, optional
|
|
:param dask: Please refer to xarray.apply_ufunc()
|
|
:type dask: {"forbidden", "allowed", "parallelized"}, optional
|
|
:return: The 'fake' background to help removing the fringes
|
|
:rtype: xarray array
|
|
"""
|
|
if not reshape is None:
|
|
self.reshape = reshape
|
|
if not referenceImages is None:
|
|
self.referenceImages = referenceImages
|
|
if not mask is None:
|
|
self.mask = mask
|
|
if self.P is None:
|
|
self.decompose_referenceImages()
|
|
|
|
if self.reshape:
|
|
absorptionImages = self._reshape_absorption_images(absorptionImages)
|
|
|
|
self.temp = self._referenceImages[self.k, :].T
|
|
|
|
optrefimages = xr.apply_ufunc(self._fringe_removal, absorptionImages, input_core_dims=[['axis']], output_core_dims=[['axis']], dask=dask, vectorize=True, output_dtypes=float)
|
|
|
|
return optrefimages.unstack() |