def proximal_numpy(self, in_arr, tau, out = None):
res = regularisers.TNV(in_arr,
self.regularisation_parameter,
self.iterationsTNV,
self.tolerance)
return res, []
def proximal_numpy(self, in_arr, tau):
if in_arr.ndim != 3:
# https://github.com/vais-ral/CCPi-Regularisation-Toolkit/blob/413c6001003c6f1272aeb43152654baaf0c8a423/src/Python/src/cpu_regularisers.pyx#L584-L588
raise ValueError('Only 3D data is supported. Passed data has {} dimensions'.format(in_arr.ndim))
res = regularisers.TNV(in_arr,
self.alpha * tau,
self.max_iteration,
self.tolerance)
return res, []
def proximal(self, x, tau, out=None):
in_arr = np.asarray(x.as_array(), dtype=np.float32, order='C')
res = regularisers.TNV(in_arr, self.regularisation_parameter,
self.iterationsTNV, self.tolerance)
if out is not None:
out.fill(res)
else:
out = x.copy()
out.fill(res)
return out
def proximal(self, x, tau, out=None):
pars = {'algorithm' : TNV, \
'input' : np.asarray(x.as_array(), dtype=np.float32),\
'regularisation_parameter':self.regularisation_parameter, \
'number_of_iterations' :self.iterationsTNV,\
'tolerance_constant':self.tolerance}
res = regularisers.TNV(pars['input'], pars['regularisation_parameter'],
pars['number_of_iterations'],
pars['tolerance_constant'])
#self.info = info
if out is not None:
out.fill(res)
else:
out = x.copy()
out.fill(res)
return out
def test_functionality_TNV(self):
# fake a 2D+channel image
d = dataexample.SYNCHROTRON_PARALLEL_BEAM_DATA.get()
ig = ImageGeometry(160, 135, channels=91)
data = ig.allocate(None)
data.fill(d)
del d
datarr = data.as_array()
from cil.plugins.ccpi_regularisation.functions import TNV
from ccpi.filters import regularisers
tau = 1.
# CIL defaults
outrgl = regularisers.TNV(datarr, 1, 100, 1e-6 )
fcil = TNV()
outcil = fcil.proximal(data, tau=tau)
np.testing.assert_almost_equal(outrgl, outcil.as_array())