class Program(object):
"""
Main class that wraps all object initializations and the subsequent
computations.
"""
def __init__(self, phantom=None):
self.__phantom = phantom
self.__projectionSimulator = None
self.__systemMatrix = None
self.__systemMatrixEvaluator = None
self.__algorithm = None
self.__sinogram = None
self.__result = None
self.__rmse = None
self.__figureIndex = 0
@property
def phantom(self):
return self.__phantom
@phantom.setter
def phantom(self, phantom):
self.__phantom = phantom
@property
def projectionSimulator(self):
return self.__projectionSimulator
@projectionSimulator.setter
def projectionSimulator(self, ps):
self.__projectionSimulator = ps
@property
def result(self):
if self.__result is None:
return self.__algorithm.result
return self.__result
@property
def rmse(self):
return self.__rmse
@property
def sinogram(self):
return self.__sinogram
@property
def systemMatrix(self):
return self.__systemMatrix
def compute(self):
"""
Starts the computation in the algorithm and stores the result.
Has become obsolete because computation is done via TaskHandler.
"""
if self.__algorithm is None:
return
self.__algorithm.compute()
self.__result = self.__algorithm.result
def computeRmse(self, groundTruth=None):
"""
Evaluates the quality of a reconstructed slice (after a certain number
of iterations) using the Root Mean Square Error w.r.t. the ground truth
phantom.
:param groundTruth | np.2darray or None
"""
if groundTruth is None:
groundTruth = self.__phantom.data
iteration = []
rmse = []
invNrPixels = 1.0/(np.prod(groundTruth.shape))
for i, image in enumerate(self.result):
if image.shape != groundTruth.shape:
print "Mismatch in data dimensions. Skipping RMSE computation nr %i.", i
continue
iteration.append(i+1)
difference = image - groundTruth
rmse.append(np.sqrt( np.sum(difference*difference) * invNrPixels ))
self.__rmse = (np.array(iteration), np.array(rmse))
def plot(self, data):
""" Plot routine for convenience. """
#plt.figure(self.__figureIndex)
self.__figureIndex += 1
plt.imshow(data, cmap="gray", interpolation="nearest")
#plt.show()
def setUp(self, views=100, start=0, stop=180, nrIter=10,
mode=Algorithm.Mode.ADDITIVE_ART):
"""
Subsequently sets up the phantom, simulated projections, system matrix.
Eventually creates the algorithm.
:param view, start, stop, nrIter | int
mode | Algorithm.Mode
"""
if self.phantom is None:
self.__phantom = Phantom(size=100)
self.__phantom.create()
if self.phantom is None:
return
if self.phantom.size is None: #error prone
return
self.__projectionSimulator = ProjectionSimulator(views,
self.phantom.size, self.phantom)
self.__projectionSimulator.initProjections()
self.__projectionSimulator.projectAll(start, stop)
self.__sinogram = self.__projectionSimulator.projections
self.__systemMatrixEvaluator = SystemMatrixEvaluator(views,
self.phantom.size, SystemMatrixEvaluator.Mode.BINARY)
self.__systemMatrixEvaluator.initSystemMatrix(self.phantom.size, views)
self.__systemMatrixEvaluator.evaluate(start, stop, self.phantom)
self.__systemMatrix = self.__systemMatrixEvaluator.systemMatrix
if self.systemMatrix is None or self.sinogram is None:
return
self.__algorithm = Algorithm(mode, self.sinogram, self.systemMatrix, nrIter)
class Program(object):
"""
Main class that wraps all object initializations and the subsequent
computations.
"""
def __init__(self, phantom=None):
self.__phantom = phantom
self.__projectionSimulator = None
self.__systemMatrix = None
self.__systemMatrixEvaluator = None
self.__algorithm = None
self.__sinogram = None
self.__result = None
self.__rmse = None
self.__figureIndex = 0
@property
def phantom(self):
return self.__phantom
@phantom.setter
def phantom(self, phantom):
self.__phantom = phantom
@property
def projectionSimulator(self):
return self.__projectionSimulator
@projectionSimulator.setter
def projectionSimulator(self, ps):
self.__projectionSimulator = ps
@property
def result(self):
if self.__result is None:
return self.__algorithm.result
return self.__result
@property
def rmse(self):
return self.__rmse
@property
def sinogram(self):
return self.__sinogram
@property
def systemMatrix(self):
return self.__systemMatrix
def compute(self):
"""
Starts the computation in the algorithm and stores the result.
Has become obsolete because computation is done via TaskHandler.
"""
if self.__algorithm is None:
return
self.__algorithm.compute()
self.__result = self.__algorithm.result
def computeRmse(self, groundTruth=None):
"""
Evaluates the quality of a reconstructed slice (after a certain number
of iterations) using the Root Mean Square Error w.r.t. the ground truth
phantom.
:param groundTruth | np.2darray or None
"""
if groundTruth is None:
groundTruth = self.__phantom.data
iteration = []
rmse = []
invNrPixels = 1.0 / (np.prod(groundTruth.shape))
for i, image in enumerate(self.result):
if image.shape != groundTruth.shape:
print "Mismatch in data dimensions. Skipping RMSE computation nr %i.", i
continue
iteration.append(i + 1)
difference = image - groundTruth
rmse.append(np.sqrt(np.sum(difference * difference) * invNrPixels))
self.__rmse = (np.array(iteration), np.array(rmse))
def plot(self, data):
""" Plot routine for convenience. """
#plt.figure(self.__figureIndex)
self.__figureIndex += 1
plt.imshow(data, cmap="gray", interpolation="nearest")
#plt.show()
def setUp(self,
views=100,
start=0,
stop=180,
nrIter=10,
mode=Algorithm.Mode.ADDITIVE_ART):
"""
Subsequently sets up the phantom, simulated projections, system matrix.
Eventually creates the algorithm.
:param view, start, stop, nrIter | int
mode | Algorithm.Mode
"""
if self.phantom is None:
self.__phantom = Phantom(size=100)
self.__phantom.create()
if self.phantom is None:
return
if self.phantom.size is None: #error prone
return
self.__projectionSimulator = ProjectionSimulator(
views, self.phantom.size, self.phantom)
self.__projectionSimulator.initProjections()
self.__projectionSimulator.projectAll(start, stop)
self.__sinogram = self.__projectionSimulator.projections
self.__systemMatrixEvaluator = SystemMatrixEvaluator(
views, self.phantom.size, SystemMatrixEvaluator.Mode.BINARY)
self.__systemMatrixEvaluator.initSystemMatrix(self.phantom.size, views)
self.__systemMatrixEvaluator.evaluate(start, stop, self.phantom)
self.__systemMatrix = self.__systemMatrixEvaluator.systemMatrix
if self.systemMatrix is None or self.sinogram is None:
return
self.__algorithm = Algorithm(mode, self.sinogram, self.systemMatrix,
nrIter)
