def lowr_comp(self, x):
"""Calculate low-rank component of the cost
This method returns the nuclear norm error of the deconvolved data in
matrix form
Parameters
----------
x : np.ndarray
Deconvolved data array
Returns
-------
float low-rank cost component
"""
x_prime = cube2matrix(x)
nuc_norm = nuclear_norm(x_prime)
if self.verbose:
print(' - NUCLEAR NORM:', nuc_norm)
return self.lambda_lowr * nuc_norm
def func(x):
HX = cube2matrix(self.H_op(x))
StSHX = self._St.dot(self._S.dot(HX))
return (self.Ht_op(self.H_op(x)) + self.lambda_reg *
self.Ht_op(matrix2cube(StSHX, x[0].shape)))
def svd_thresh_coef(data, operator, threshold, thresh_type='hard'):
"""Threshold the singular values coefficients
This method thresholds the input data using singular value decomposition
Parameters
----------
data : np.ndarray
Input data array
operator : class
Operator class instance
threshold : float, optional
Threshold value
threshold_type : str {'hard', 'soft'}
Type of noise to be added (default is 'hard')
Returns
-------
np.ndarray thresholded data
Raises
------
ValueError
For invalid string entry for n_pc
"""
# Convert data cube to matrix.
data_matrix = cube2matrix(data)
# Get SVD of data matrix.
u, s, v = np.linalg.svd(data_matrix, full_matrices=False)
# Compute coefficients.
a = np.dot(np.diag(s), v)
# Compute threshold matrix.
u_cube = matrix2cube(u, data.shape[1:])
ti = np.array([np.linalg.norm(x) for x in operator(u_cube)])
ti = np.repeat(ti, a.shape[1]).reshape(a.shape)
threshold *= ti
# Remove noise from coefficients.
a_new = thresh(a, threshold, thresh_type)
# Return the thresholded image.
return np.dot(u, a_new)
def _calc_shape_grad(self, x):
"""Get the gradient of the shape constraint component
This method calculates the gradient value of the shape constraint
component
Parameters
----------
x : np.ndarray
Input data array, an array of recovered 2D images
"""
HXY = cube2matrix(self.H_op(x) - self._y)
StSHXY = self._St.dot(self._S.dot(HXY))
return self.Ht_op(matrix2cube(StSHXY, x[0].shape))
def __init__(self, data, psf, psf_type='fixed', lambda_reg=1):
self.grad_type = 'shape'
self._y = np.copy(data)
self._psf = np.copy(psf)
self._psf_type = psf_type
self.lambda_reg = lambda_reg
self._St = cube2matrix(shape_project(data.shape[1:]))
self._S = self._St.T
def func(x):
HX = cube2matrix(self.H_op(x))
StSHX = self._St.dot(self._S.dot(HX))
return (self.Ht_op(self.H_op(x)) + self.lambda_reg *
self.Ht_op(matrix2cube(StSHX, x[0].shape)))
PowerMethod.__init__(self, func, self._y.shape)