def cg_step(self, rhs, m, uker, n_iter):
FhFWm = self.do_FhWFm(
m * self.st['sensemap']) * self.st['sensemap'].conj()
FhFWm = CsSolver.CombineMulti(FhFWm, -1)
lapla_m = self.do_laplacian(m, uker)
lhs = FhFWm - self.LMBD * lapla_m + 2.0 * self.gamma * m
#C_m= lhs - rhs
r = rhs - lhs
p = r
for pp in range(0, n_iter):
Ap = self.do_FhWFm(
p * self.st['sensemap']) * self.st['sensemap'].conj()
Ap = CsSolver.CombineMulti(Ap, -1)
Ap = Ap - self.LMBD * self.do_laplacian(
p, uker) + 2.0 * self.gamma * p
upper_ratio = numpy.sum((r.conj() * r)[:])
lower_ratio = numpy.sum((p.conj() * Ap)[:])
alfa_k = upper_ratio / lower_ratio
print('r', upper_ratio, 'alpha_k', alfa_k)
#alfa_k = 0.3
m = m + alfa_k * p
r2 = r - alfa_k * Ap
beta_k = numpy.sum((r2.conj() * r2)[:]) / numpy.sum(
(r.conj() * r)[:])
r = r2
p = r + beta_k * p
return m
def external_update(self, u, f, uf, f0,
u0): # overload the update function
CsSolver.checkmax(self.st['sensemap'])
tmpuf = u * self.st['sensemap']
tmpuf = numpy.transpose(tmpuf, (1, 2, 3, 0))
tmp_shape = tmpuf.shape
tmpuf = numpy.reshape(tmpuf,
tmp_shape[0:2] + (numpy.prod(tmp_shape[2:4]), ),
order='F')
tmpuf = self.CsTransform.forwardbackward(tmpuf)
tmpuf = numpy.reshape(tmpuf, tmp_shape, order='F')
tmpuf = numpy.transpose(tmpuf, (3, 0, 1, 2))
tmpuf = tmpuf * self.st['sensemap'].conj()
# tmpuf=self.st['sensemap'].conj()*(
# self.CsTransform.forwardbackward(
# u*self.st['sensemap']))
if self.st['senseflag'] == 1:
tmpuf = CsSolver.CombineMulti(tmpuf, -1)
print('start of ext_update')
# checkmax(u)
# checkmax(tmpuf)
# checkmax(self.u0)
# checkmax(uf)
fact = numpy.sum((self.u0 - tmpuf)**2) / numpy.sum((u0)**2)
fact = numpy.abs(fact.real)
fact = numpy.sqrt(fact)
print('fact', fact)
# fact=1.0/(1.0+numpy.exp(-(fact-0.5)*self.thresh_scale))
tmpuf = CsSolver.Normalize(tmpuf) * numpy.max(numpy.abs(u0[:]))
uf = uf + (u0 - tmpuf) * 1.0 #*fact
uf = CsSolver.Normalize(uf) * numpy.max(numpy.abs(u0[:]))
CsSolver.checkmax(tmpuf)
CsSolver.checkmax(u0)
CsSolver.checkmax(uf)
# for jj in range(0,u.shape[-1]):
# u[...,jj] = u[...,jj]*self.st['sn']# rescale the final image intensity
print('end of ext_update')
murf = uf
return (f, uf, murf, u)
def kernel(self, f_internal, st, mu, LMBD, gamma, nInner, nBreg):
self.st['sensemap'] = self.st['sensemap'] * self.st['mask']
tse = self.f.tse
# tse=numpy.abs(numpy.mean(self.st['sensemap'],-1))
tse = CsSolver.appendmat(tse, self.st['Nd'][1])
#tse=Normalize(tse)
tse = numpy.transpose(tse, (0, 1, 3, 2))
self.ttse = CsSolver.Normalize(tse)
self.tse0 = CsSolver.CombineMulti(tse, -1)
self.filter = numpy.ones(tse.shape)
dpss = numpy.kaiser(tse.shape[1], 1.0) * 10.0
for ppp in range(0, tse.shape[1]):
self.filter[:, ppp, :, :] = self.filter[:, ppp, :, :] * dpss[ppp]
print('tse.shape', tse.shape)
# L= numpy.size(f)/st['M']
# image_dim=st['Nd']+(L,)
#
# if numpy.ndim(f) == 1:# preventing row vector
# f=numpy.reshape(f,(numpy.shape(f)[0],1),order='F')
# f0 = numpy.copy(f) # deep copy to prevent scope f0 to f
## u = numpy.zeros(image_dim,dtype=numpy.complex64)
f0 = numpy.copy(f_internal)
f = numpy.copy(f_internal)
# u0=self.data2rho(f_internal,
# self.f.dim_x,
# self.st['Nd'][0],
# self.st['Nd'][1],
# self.f.ncoils,
# self.CsTransform
# ) # doing spatial transform
u0 = self.fun1(f_internal)
pdf = self.f.pdf
pdf = CsSolver.appendmat(pdf, self.st['Nd'][1])
pdf = numpy.transpose(pdf, (0, 1, 3, 2))
# u0 = scipy.fftpack.fftn(u0,axes=(1,))
# u0 = scipy.fftpack.fftshift(u0,axes=(1,))
# #u0[:,:,u0.shape[2]/2,:] = u0[:,:,u0.shape[2]/2,:]/pdf[:,:,u0.shape[2]/2,:]
# u0 = u0#/pdf
# u0 = scipy.fftpack.ifftshift(u0,axes=(1,))
# u0 = scipy.fftpack.ifftn(u0,axes=(1,))
# print('self.f.pdf.shape',self.f.pdf.shape)
# for pj in range(0,4):
# matplotlib.pyplot.imshow(self.f.pdf[:,:,pj].real)
# matplotlib.pyplot.show()
u0 = self.fun2(u0)
u0 = self.fun3(u0)
u0 = u0 * self.st['sensemap'].conj()
u0 = CsSolver.CombineMulti(u0, -1)
#u0 = u0*self.filter
uker = self.create_laplacian_kernel()
uker = CsSolver.appendmat(uker, u0.shape[3])
self.u0 = u0
u = numpy.copy(self.tse0)
print('u0.shape', u0.shape)
(xx, bb, dd) = self.make_split_variables(u)
uf = numpy.copy(u) # only used for ISRA, written here for generality
murf = numpy.copy(u) # initial values
# #===============================================================================
#u_stack = numpy.empty(st['Nd']+(nBreg,),dtype=numpy.complex)
for outer in numpy.arange(0, nBreg):
for inner in numpy.arange(0, nInner):
# update u
print('iterating', [inner, outer])
#===============================================================
# update u # simple k-space deconvolution to guess initial u
u = self.update_u(murf, u, uker, xx, bb)
c = numpy.max(numpy.abs(u[:])) # Rough coefficient
# to correct threshold of nonlinear shrink
#===================================================================
# # update d
#===================================================================
#===================================================================
# Shrinkage: remove tiny values "in somewhere sparse!"
# dx+bx should be sparse!
#===================================================================
# shrinkage
#===================================================================
dd = self.update_d(u, dd)
xx = self.shrink(
dd, bb, c * 1.0 / LMBD / numpy.sqrt(numpy.prod(st['Nd'])))
#===============================================================
#===================================================================
# # update b
#===================================================================
bb = self.update_b(bb, dd, xx)
# if outer < nBreg: # do not update in the last loop
(f, uf, murf,
u) = self.external_update(u, f, uf, f0,
u0) # update outer Split_bregman
u = CsSolver.Normalize(u)
for pp in range(0, u0.shape[2]):
matplotlib.pyplot.subplot(
numpy.sqrt(u0.shape[2]) + 1,
numpy.sqrt(u0.shape[2]) + 1, pp)
matplotlib.pyplot.imshow(numpy.sum(numpy.abs(u[..., pp, :]), -1),
norm=norm,
interpolation='nearest')
matplotlib.pyplot.show()
#
return (u, uf)
