def upsample(self, new_domain_forward, new_domain_backward):
r'''
Upsamples the displacement fields and scales the affine
pre- and post-multiplication affine matrices by a factor of 2. The
final outcome is that this transformation can be used in an upsampled
domain.
'''
if self.dim == 2:
if self.forward != None:
self.forward = 2*np.array(
tf.upsample_displacement_field(
self.forward,
np.array(new_domain_forward).astype(np.int32)))
if self.backward != None:
self.backward = 2*np.array(
tf.upsample_displacement_field(
self.backward,
np.array(new_domain_backward).astype(np.int32)))
else:
if self.forward != None:
self.forward = 2*np.array(
tf.upsample_displacement_field3D(
self.forward,
np.array(new_domain_forward).astype(np.int32)))
if self.backward != None:
self.backward = 2*np.array(
tf.upsample_displacement_field3D(
self.backward,
np.array(new_domain_backward).astype(np.int32)))
self.scale_affines(2.0)
def estimateMultimodalDiffeomorphicField3DMultiScale(movingPyramid,
fixedPyramid,
initAffine,
lambdaParam,
maxOuterIter,
level=0,
displacementList=None):
n = len(movingPyramid)
quantizationLevels = 256
if (level == (n - 1)):
displacement = estimateNewMultimodalDiffeomorphicField3D(
movingPyramid[level], fixedPyramid[level], initAffine, lambdaParam,
quantizationLevels, maxOuterIter[level], None, level == 0)
if (displacementList != None):
displacementList.insert(0, displacement)
return displacement
subAffine = initAffine.copy()
#subAffine=initAffine.copy()*0.5
subAffine[:3, 3] *= 0.5
subDisplacement = estimateMultimodalDiffeomorphicField3DMultiScale(
movingPyramid, fixedPyramid, subAffine, lambdaParam, maxOuterIter,
level + 1, displacementList)
sh = np.array(fixedPyramid[level].shape).astype(np.int32)
upsampled = np.array(tf.upsample_displacement_field3D(subDisplacement,
sh)) * 2
newDisplacement = estimateNewMultimodalDiffeomorphicField3D(
movingPyramid[level], fixedPyramid[level], initAffine, lambdaParam,
quantizationLevels, maxOuterIter[level], upsampled, level == 0)
newDisplacement += upsampled
if (displacementList != None):
displacementList.insert(0, newDisplacement)
return newDisplacement
def estimateMonomodalDiffeomorphicField3DMultiScale(movingPyramid,
fixedPyramid, lambdaParam,
maxOuterIter, level,
displacementList):
n = len(movingPyramid)
if (level == (n - 1)):
displacement = estimateNewMonomodalDiffeomorphicField3D(
movingPyramid[level], fixedPyramid[level], lambdaParam,
maxOuterIter[level], None, level == 0)
if (displacementList != None):
displacementList.insert(0, displacement)
return displacement
subDisplacement = estimateMonomodalDiffeomorphicField3DMultiScale(
movingPyramid, fixedPyramid, lambdaParam, maxOuterIter, level + 1,
displacementList)
sh = np.array(movingPyramid[level].shape)
upsampled = np.array(tf.upsample_displacement_field3D(subDisplacement,
sh)) * 2
newDisplacement = estimateNewMonomodalDiffeomorphicField3D(
movingPyramid[level], fixedPyramid[level], lambdaParam,
maxOuterIter[level], upsampled, level == 0)
newDisplacement += upsampled
if (displacementList != None):
displacementList.insert(0, newDisplacement)
return np.array(newDisplacement)
def estimateMultimodalSyN3DMultiScale(movingPyramid,
fixedPyramid,
initAffine,
lambdaParam,
maxOuterIter,
level=0):
n = len(movingPyramid)
quantizationLevels = 256
if (level == (n - 1)):
totalF, totalFInv, totalM, totalMInv = estimateNewMultimodalSyNField3D(
movingPyramid[level], fixedPyramid[level], None, None, None, None,
initAffine, lambdaParam, quantizationLevels, maxOuterIter[level],
level == 0)
return totalF, totalFInv, totalM, totalMInv
subAffine = initAffine.copy()
subAffine[:3, 3] *= 0.5
subF, subFInv, subM, subMInv = estimateMultimodalSyN3DMultiScale(
movingPyramid, fixedPyramid, subAffine, lambdaParam, maxOuterIter,
level + 1)
sh = np.array(fixedPyramid[level].shape).astype(np.int32)
upF = np.array(tf.upsample_displacement_field3D(subF, sh)) * 2
upFInv = np.array(tf.upsample_displacement_field3D(subFInv, sh)) * 2
upM = np.array(tf.upsample_displacement_field3D(subM, sh)) * 2
upMInv = np.array(tf.upsample_displacement_field3D(subMInv, sh)) * 2
del subF
del subFInv
del subM
del subMInv
totalF, totalFInv, totalM, totalMInv = estimateNewMultimodalSyNField3D(
movingPyramid[level], fixedPyramid[level], upF, upFInv, upM, upMInv,
initAffine, lambdaParam, quantizationLevels, maxOuterIter[level],
level == 0)
if level == 0:
totalF = np.array(tf.compose_vector_fields3D(
totalF, totalMInv)) #Multiply bw by 0.5??
totalM = np.array(tf.compose_vector_fields3D(
totalM, totalFInv)) #Multiply bw by 0.5??
return totalM, totalF
return totalF, totalFInv, totalM, totalMInv
def estimateMultimodalSyN3DMultiScale(movingPyramid, fixedPyramid, initAffine, lambdaParam, maxOuterIter, level=0):
n=len(movingPyramid)
quantizationLevels=256
if(level==(n-1)):
totalF, totalFInv, totalM, totalMInv=estimateNewMultimodalSyNField3D(movingPyramid[level], fixedPyramid[level], None, None, None, None, initAffine, lambdaParam, quantizationLevels, maxOuterIter[level], level==0)
return totalF, totalFInv, totalM, totalMInv
subAffine=initAffine.copy()
subAffine[:3,3]*=0.5
subF, subFInv, subM, subMInv=estimateMultimodalSyN3DMultiScale(movingPyramid, fixedPyramid, subAffine, lambdaParam, maxOuterIter, level+1)
sh=np.array(fixedPyramid[level].shape).astype(np.int32)
upF=np.array(tf.upsample_displacement_field3D(subF, sh))*2
upFInv=np.array(tf.upsample_displacement_field3D(subFInv, sh))*2
upM=np.array(tf.upsample_displacement_field3D(subM, sh))*2
upMInv=np.array(tf.upsample_displacement_field3D(subMInv, sh))*2
del subF
del subFInv
del subM
del subMInv
totalF, totalFInv, totalM, totalMInv=estimateNewMultimodalSyNField3D(movingPyramid[level], fixedPyramid[level], upF, upFInv, upM, upMInv, initAffine, lambdaParam, quantizationLevels, maxOuterIter[level], level==0)
if level==0:
totalF=np.array(tf.compose_vector_fields3D(totalF, totalMInv))#Multiply bw by 0.5??
totalM=np.array(tf.compose_vector_fields3D(totalM, totalFInv))#Multiply bw by 0.5??
return totalM, totalF
return totalF, totalFInv, totalM, totalMInv
def estimateMonomodalDiffeomorphicField3DMultiScale(movingPyramid, fixedPyramid, lambdaParam, maxOuterIter, level, displacementList):
n=len(movingPyramid)
if(level==(n-1)):
displacement=estimateNewMonomodalDiffeomorphicField3D(movingPyramid[level], fixedPyramid[level], lambdaParam, maxOuterIter[level], None, level==0)
if(displacementList!=None):
displacementList.insert(0,displacement)
return displacement
subDisplacement=estimateMonomodalDiffeomorphicField3DMultiScale(movingPyramid, fixedPyramid, lambdaParam, maxOuterIter, level+1, displacementList)
sh=np.array(movingPyramid[level].shape)
upsampled=np.array(tf.upsample_displacement_field3D(subDisplacement, sh))*2
newDisplacement=estimateNewMonomodalDiffeomorphicField3D(movingPyramid[level], fixedPyramid[level], lambdaParam, maxOuterIter[level], upsampled, level==0)
newDisplacement+=upsampled
if(displacementList!=None):
displacementList.insert(0, newDisplacement)
return np.array(newDisplacement)
def estimateMultimodalDiffeomorphicField3DMultiScale(movingPyramid, fixedPyramid, initAffine, lambdaParam, maxOuterIter, level=0, displacementList=None):
n=len(movingPyramid)
quantizationLevels=256
if(level==(n-1)):
displacement=estimateNewMultimodalDiffeomorphicField3D(movingPyramid[level], fixedPyramid[level], initAffine, lambdaParam, quantizationLevels, maxOuterIter[level], None, level==0)
if(displacementList!=None):
displacementList.insert(0, displacement)
return displacement
subAffine=initAffine.copy()
#subAffine=initAffine.copy()*0.5
subAffine[:3,3]*=0.5
subDisplacement=estimateMultimodalDiffeomorphicField3DMultiScale(movingPyramid, fixedPyramid, subAffine, lambdaParam, maxOuterIter, level+1, displacementList)
sh=np.array(fixedPyramid[level].shape).astype(np.int32)
upsampled=np.array(tf.upsample_displacement_field3D(subDisplacement, sh))*2
newDisplacement=estimateNewMultimodalDiffeomorphicField3D(movingPyramid[level], fixedPyramid[level], initAffine, lambdaParam, quantizationLevels, maxOuterIter[level], upsampled, level==0)
newDisplacement+=upsampled
if(displacementList!=None):
displacementList.insert(0, newDisplacement)
return newDisplacement
def w_cycle_3d(n, k, delta_field, sigma_field, gradient_field, target,
lambda_param, displacement, depth = 0):
r'''
Multi-resolution Gauss-Seidel solver: solves the linear system by performing
two v-cycles at each resolution, which corresponds to the w-cycle scheme
proposed by Bruhn and Weickert[1].
[1] Andres Bruhn and Joachim Weickert, "Towards ultimate motion estimation:
combining highest accuracy with real-time performance",
10th IEEE International Conference on Computer Vision, 2005. ICCV 2005.
'''
#presmoothing
for i in range(k):
error = tf.iterate_residual_displacement_field_SSD3D(delta_field,
sigma_field,
gradient_field,
target,
lambda_param,
displacement)
if printEnergy and depth == 0:
energy = tf.compute_energy_SSD3D(delta_field,
sigma_field,
gradient_field,
lambda_param,
displacement)
print 'Energy after top-level iter', i+1, ' [first]:', energy
if n == 0:
return error
residual = tf.compute_residual_displacement_field_SSD3D(delta_field,
sigma_field,
gradient_field,
target,
lambda_param,
displacement,
None)
sub_residual = np.array(tf.downsample_displacement_field3D(residual))
del residual
#solve at coarcer grid
subsigma_field = None
if sigma_field != None:
subsigma_field = tf.downsample_scalar_field3D(sigma_field)
subdelta_field = tf.downsample_scalar_field3D(delta_field)
subgradient_field = np.array(
tf.downsample_displacement_field3D(gradient_field))
shape = np.array(displacement.shape).astype(np.int32)
sub_displacement = np.zeros(
shape = ((shape[0]+1)//2, (shape[1]+1)//2, (shape[2]+1)//2, 3 ),
dtype = np.float64)
sublambda_param = lambda_param*0.25
w_cycle_3d(n-1, k, subdelta_field, subsigma_field, subgradient_field,
sub_residual, sublambda_param, sub_displacement, depth+1)
displacement += np.array(
tf.upsample_displacement_field3D(sub_displacement, shape))
if printEnergy and depth == 0:
energy = tf.compute_energy_SSD3D(delta_field, sigma_field,
gradient_field, lambda_param,
displacement)
print 'Energy after low-res iteration[first]:', energy
#post-smoothing (second smoothing)
for i in range(k):
error = tf.iterate_residual_displacement_field_SSD3D(delta_field,
sigma_field,
gradient_field,
target,
lambda_param,
displacement)
if printEnergy and depth == 0:
energy = tf.compute_energy_SSD3D(delta_field,
sigma_field,
gradient_field,
lambda_param,
displacement)
print 'Energy after top-level iter', i+1, ' [second]:', energy
residual = tf.compute_residual_displacement_field_SSD3D(delta_field,
sigma_field,
gradient_field,
target,
lambda_param,
displacement,
None)
sub_residual = np.array(tf.downsample_displacement_field3D(residual))
del residual
sub_displacement[...] = 0
w_cycle_3d(n-1, k, subdelta_field, subsigma_field, subgradient_field,
sub_residual, sublambda_param, sub_displacement, depth+1)
displacement += np.array(
tf.upsample_displacement_field3D(sub_displacement, shape))
if printEnergy and depth == 0:
energy = tf.compute_energy_SSD3D(delta_field, sigma_field,
gradient_field, lambda_param,
displacement)
print 'Energy after low-res iteration[second]:', energy
for i in range(k):
error = tf.iterate_residual_displacement_field_SSD3D(delta_field,
sigma_field,
gradient_field,
target,
lambda_param,
displacement)
if printEnergy and depth == 0:
energy = tf.compute_energy_SSD3D(delta_field,
sigma_field,
gradient_field,
lambda_param,
displacement)
print 'Energy after top-level iter', i+1, ' [third]:', energy
try:
energy
except NameError:
energy = tf.compute_energy_SSD3D(delta_field, sigma_field,
gradient_field, lambda_param,
displacement)
return energy
def single_cycle_3d(n, k, delta_field, sigma_field, gradient_field,
lambda_param, displacement, depth = 0):
r'''
One-pass multi-resolution Gauss-Seidel solver: solves the SSD-like linear
system starting at the coarcer resolution and then refining at the finer
resolution.
'''
if n == 0:
for i in range(k):
error = tf.iterate_residual_displacement_field_SSD3D(delta_field,
sigma_field,
gradient_field,
None,
lambda_param,
displacement)
if printEnergy and depth == 0:
energy = tf.compute_energy_SSD3D(delta_field,
sigma_field,
gradient_field,
lambda_param,
displacement)
print 'Energy after top-level iter', i+1, ' [unique]:', energy
return error
#solve at coarcer grid
subsigma_field = None
if sigma_field != None:
subsigma_field = tf.downsample_scalar_field3D(sigma_field)
subdelta_field = tf.downsample_scalar_field3D(delta_field)
subgradient_field = np.array(
tf.downsample_displacement_field3D(gradient_field))
shape = np.array(displacement.shape).astype(np.int32)
sub_displacement = np.zeros(
shape = ((shape[0]+1)//2, (shape[1]+1)//2, (shape[2]+1)//2, 3 ),
dtype = np.float64)
sublambda_param = lambda_param*0.25
single_cycle_3d(n-1, k, subdelta_field, subsigma_field, subgradient_field,
sublambda_param, sub_displacement, depth+1)
displacement += np.array(
tf.upsample_displacement_field3D(sub_displacement, shape))
if printEnergy and depth == 0:
energy = tf.compute_energy_SSD3D(delta_field, sigma_field,
gradient_field, lambda_param,
displacement)
print 'Energy after low-res iteration:', energy
#post-smoothing
for i in range(k):
error = tf.iterate_residual_displacement_field_SSD3D(delta_field,
sigma_field,
gradient_field,
None,
lambda_param,
displacement)
if printEnergy and depth == 0:
energy = tf.compute_energy_SSD3D(delta_field,
sigma_field,
gradient_field,
lambda_param,
displacement)
print 'Energy after top-level iter', i+1, ' [unique]:', energy
try:
energy
except NameError:
energy = tf.compute_energy_SSD3D(delta_field, sigma_field,
gradient_field, lambda_param,
displacement)
return energy
def w_cycle_3d(n,
k,
delta_field,
sigma_field,
gradient_field,
target,
lambda_param,
displacement,
depth=0):
r'''
Multi-resolution Gauss-Seidel solver: solves the linear system by performing
two v-cycles at each resolution, which corresponds to the w-cycle scheme
proposed by Bruhn and Weickert[1].
[1] Andres Bruhn and Joachim Weickert, "Towards ultimate motion estimation:
combining highest accuracy with real-time performance",
10th IEEE International Conference on Computer Vision, 2005. ICCV 2005.
'''
#presmoothing
for i in range(k):
error = tf.iterate_residual_displacement_field_SSD3D(
delta_field, sigma_field, gradient_field, target, lambda_param,
displacement)
if printEnergy and depth == 0:
energy = tf.compute_energy_SSD3D(delta_field, sigma_field,
gradient_field, lambda_param,
displacement)
print 'Energy after top-level iter', i + 1, ' [first]:', energy
if n == 0:
return error
residual = tf.compute_residual_displacement_field_SSD3D(
delta_field, sigma_field, gradient_field, target, lambda_param,
displacement, None)
sub_residual = np.array(tf.downsample_displacement_field3D(residual))
del residual
#solve at coarcer grid
subsigma_field = None
if sigma_field != None:
subsigma_field = tf.downsample_scalar_field3D(sigma_field)
subdelta_field = tf.downsample_scalar_field3D(delta_field)
subgradient_field = np.array(
tf.downsample_displacement_field3D(gradient_field))
shape = np.array(displacement.shape).astype(np.int32)
sub_displacement = np.zeros(shape=((shape[0] + 1) // 2,
(shape[1] + 1) // 2,
(shape[2] + 1) // 2, 3),
dtype=np.float64)
sublambda_param = lambda_param * 0.25
w_cycle_3d(n - 1, k, subdelta_field, subsigma_field, subgradient_field,
sub_residual, sublambda_param, sub_displacement, depth + 1)
displacement += np.array(
tf.upsample_displacement_field3D(sub_displacement, shape))
if printEnergy and depth == 0:
energy = tf.compute_energy_SSD3D(delta_field, sigma_field,
gradient_field, lambda_param,
displacement)
print 'Energy after low-res iteration[first]:', energy
#post-smoothing (second smoothing)
for i in range(k):
error = tf.iterate_residual_displacement_field_SSD3D(
delta_field, sigma_field, gradient_field, target, lambda_param,
displacement)
if printEnergy and depth == 0:
energy = tf.compute_energy_SSD3D(delta_field, sigma_field,
gradient_field, lambda_param,
displacement)
print 'Energy after top-level iter', i + 1, ' [second]:', energy
residual = tf.compute_residual_displacement_field_SSD3D(
delta_field, sigma_field, gradient_field, target, lambda_param,
displacement, None)
sub_residual = np.array(tf.downsample_displacement_field3D(residual))
del residual
sub_displacement[...] = 0
w_cycle_3d(n - 1, k, subdelta_field, subsigma_field, subgradient_field,
sub_residual, sublambda_param, sub_displacement, depth + 1)
displacement += np.array(
tf.upsample_displacement_field3D(sub_displacement, shape))
if printEnergy and depth == 0:
energy = tf.compute_energy_SSD3D(delta_field, sigma_field,
gradient_field, lambda_param,
displacement)
print 'Energy after low-res iteration[second]:', energy
for i in range(k):
error = tf.iterate_residual_displacement_field_SSD3D(
delta_field, sigma_field, gradient_field, target, lambda_param,
displacement)
if printEnergy and depth == 0:
energy = tf.compute_energy_SSD3D(delta_field, sigma_field,
gradient_field, lambda_param,
displacement)
print 'Energy after top-level iter', i + 1, ' [third]:', energy
try:
energy
except NameError:
energy = tf.compute_energy_SSD3D(delta_field, sigma_field,
gradient_field, lambda_param,
displacement)
return energy
def single_cycle_3d(n,
k,
delta_field,
sigma_field,
gradient_field,
lambda_param,
displacement,
depth=0):
r'''
One-pass multi-resolution Gauss-Seidel solver: solves the SSD-like linear
system starting at the coarcer resolution and then refining at the finer
resolution.
'''
if n == 0:
for i in range(k):
error = tf.iterate_residual_displacement_field_SSD3D(
delta_field, sigma_field, gradient_field, None, lambda_param,
displacement)
if printEnergy and depth == 0:
energy = tf.compute_energy_SSD3D(delta_field, sigma_field,
gradient_field, lambda_param,
displacement)
print 'Energy after top-level iter', i + 1, ' [unique]:', energy
return error
#solve at coarcer grid
subsigma_field = None
if sigma_field != None:
subsigma_field = tf.downsample_scalar_field3D(sigma_field)
subdelta_field = tf.downsample_scalar_field3D(delta_field)
subgradient_field = np.array(
tf.downsample_displacement_field3D(gradient_field))
shape = np.array(displacement.shape).astype(np.int32)
sub_displacement = np.zeros(shape=((shape[0] + 1) // 2,
(shape[1] + 1) // 2,
(shape[2] + 1) // 2, 3),
dtype=np.float64)
sublambda_param = lambda_param * 0.25
single_cycle_3d(n - 1, k, subdelta_field, subsigma_field,
subgradient_field, sublambda_param, sub_displacement,
depth + 1)
displacement += np.array(
tf.upsample_displacement_field3D(sub_displacement, shape))
if printEnergy and depth == 0:
energy = tf.compute_energy_SSD3D(delta_field, sigma_field,
gradient_field, lambda_param,
displacement)
print 'Energy after low-res iteration:', energy
#post-smoothing
for i in range(k):
error = tf.iterate_residual_displacement_field_SSD3D(
delta_field, sigma_field, gradient_field, None, lambda_param,
displacement)
if printEnergy and depth == 0:
energy = tf.compute_energy_SSD3D(delta_field, sigma_field,
gradient_field, lambda_param,
displacement)
print 'Energy after top-level iter', i + 1, ' [unique]:', energy
try:
energy
except NameError:
energy = tf.compute_energy_SSD3D(delta_field, sigma_field,
gradient_field, lambda_param,
displacement)
return energy
