plt.figure()
plt.imshow(RecFISTA_TV_hub, vmin=0, vmax=0.2, cmap="gray")
plt.title('FISTA-PWLS-HUBER-OS-TV reconstruction')
plt.show()
#%%
from ccpi.filters.regularisers import PatchSelect
print("Pre-calculating weights for non-local patches...")
pars = {'algorithm' : PatchSelect, \
'input' : RecCGLS,\
'searchwindow': 7, \
'patchwindow': 2,\
'neighbours' : 13 ,\
'edge_parameter':0.8}
H_i, H_j, Weights = PatchSelect(pars['input'], pars['searchwindow'],
pars['patchwindow'], pars['neighbours'],
pars['edge_parameter'], 'gpu')
plt.figure()
plt.imshow(Weights[0, :, :], vmin=0, vmax=1, cmap="gray")
plt.colorbar(ticks=[0, 0.5, 1], orientation='vertical')
#%%
print("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
print("Reconstructing with FISTA PWLS-OS-NLTV method %%%%%%%%%%%%%%")
print("%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%")
_regularisation_ = {'method' : 'NLTV',
'regul_param' :0.0005,
'iterations' : 5,
'NLTV_H_i' : H_i,\
'NLTV_H_j' : H_j,\
'NLTV_Weights' : Weights,\
def process_frames(self, data):
import numpy as np
#input_temp = data[0]
#indices = np.where(np.isnan(input_temp))
#input_temp[indices] = 0.0
#self.pars['input'] = input_temp
input_temp = np.nan_to_num(data[0])
input_temp[input_temp > 10**15] = 0.0
self.pars['input'] = input_temp
#self.pars['input'] = np.nan_to_num(data[0])
# Running Ccpi-RGLTK modules on GPU
if (self.parameters['method'] == 'ROF_TV'):
(im_res,
infogpu) = ROF_TV(self.pars['input'],
self.pars['regularisation_parameter'],
self.pars['number_of_iterations'],
self.pars['time_marching_parameter'],
self.pars['tolerance_constant'], self.device)
if (self.parameters['method'] == 'FGP_TV'):
(im_res, infogpu) = FGP_TV(self.pars['input'],
self.pars['regularisation_parameter'],
self.pars['number_of_iterations'],
self.pars['tolerance_constant'],
self.pars['methodTV'],
self.pars['nonneg'], self.device)
if (self.parameters['method'] == 'SB_TV'):
(im_res, infogpu) = SB_TV(self.pars['input'],
self.pars['regularisation_parameter'],
self.pars['number_of_iterations'],
self.pars['tolerance_constant'],
self.pars['methodTV'], self.device)
if (self.parameters['method'] == 'TGV'):
(im_res,
infogpu) = TGV(self.pars['input'],
self.pars['regularisation_parameter'],
self.pars['alpha1'], self.pars['alpha0'],
self.pars['number_of_iterations'],
self.pars['LipshitzConstant'],
self.pars['tolerance_constant'], self.device)
if (self.parameters['method'] == 'LLT_ROF'):
(im_res,
infogpu) = LLT_ROF(self.pars['input'],
self.pars['regularisation_parameter'],
self.pars['regularisation_parameterLLT'],
self.pars['number_of_iterations'],
self.pars['time_marching_parameter'],
self.pars['tolerance_constant'], self.device)
if (self.parameters['method'] == 'NDF'):
(im_res, infogpu) = NDF(
self.pars['input'], self.pars['regularisation_parameter'],
self.pars['edge_parameter'], self.pars['number_of_iterations'],
self.pars['time_marching_parameter'],
self.pars['penalty_type'], self.pars['tolerance_constant'],
self.device)
if (self.parameters['method'] == 'DIFF4th'):
(im_res,
infogpu) = Diff4th(self.pars['input'],
self.pars['regularisation_parameter'],
self.pars['edge_parameter'],
self.pars['number_of_iterations'],
self.pars['time_marching_parameter'],
self.pars['tolerance_constant'], self.device)
if (self.parameters['method'] == 'NLTV'):
pars_NLTV = {'algorithm' : PatchSelect, \
'input' : self.pars['input'],\
'searchwindow': 9, \
'patchwindow': 2,\
'neighbours' : 17 ,\
'edge_parameter':self.pars['edge_parameter']}
H_i, H_j, Weights = PatchSelect(pars_NLTV['input'],
pars_NLTV['searchwindow'],
pars_NLTV['patchwindow'],
pars_NLTV['neighbours'],
pars_NLTV['edge_parameter'],
self.device)
parsNLTV_init = {'algorithm' : NLTV, \
'input' : pars_NLTV['input'],\
'H_i': H_i, \
'H_j': H_j,\
'H_k' : 0,\
'Weights' : Weights,\
'regularisation_parameter': self.pars['regularisation_parameter'],\
'iterations': self.pars['number_of_iterations']}
im_res = NLTV(parsNLTV_init['input'], parsNLTV_init['H_i'],
parsNLTV_init['H_j'], parsNLTV_init['H_k'],
parsNLTV_init['Weights'],
parsNLTV_init['regularisation_parameter'],
parsNLTV_init['iterations'])
del H_i, H_j, Weights
#print "calling process frames", data[0].shape
return im_res
