def augment_images_intensity(image_list, output_prefix, output_suffix):
'''
Generate intensity modified images from the originals.
Args:
image_list (iterable containing SimpleITK images): The images which we whose intensities we modify.
output_prefix (string): output file name prefix (file name: output_prefixi_FilterName.output_suffix).
output_suffix (string): output file name suffix (file name: output_prefixi_FilterName.output_suffix).
Example in use: intensity_augmened_image = augment_images_intensity(data, os.path.join(OUTPUT_DIR, 'intensity_aug'), 'mha')
'''
# Create a list of intensity modifying filters, which we apply to the given images
filter_list = []
# Smoothing filters
filter_list.append(sitk.SmoothingRecursiveGaussianImageFilter())
filter_list[-1].SetSigma(2.0)
filter_list.append(sitk.DiscreteGaussianImageFilter())
filter_list[-1].SetVariance(4.0)
filter_list.append(sitk.BilateralImageFilter())
filter_list[-1].SetDomainSigma(4.0)
filter_list[-1].SetRangeSigma(8.0)
filter_list.append(sitk.MedianImageFilter())
filter_list[-1].SetRadius(8)
# Noise filters using default settings
# Filter control via SetMean, SetStandardDeviation.
filter_list.append(sitk.AdditiveGaussianNoiseImageFilter())
# Filter control via SetProbability
filter_list.append(sitk.SaltAndPepperNoiseImageFilter())
# Filter control via SetScale
filter_list.append(sitk.ShotNoiseImageFilter())
# Filter control via SetStandardDeviation
filter_list.append(sitk.SpeckleNoiseImageFilter())
filter_list.append(sitk.AdaptiveHistogramEqualizationImageFilter())
filter_list[-1].SetAlpha(1.0)
filter_list[-1].SetBeta(0.0)
filter_list.append(sitk.AdaptiveHistogramEqualizationImageFilter())
filter_list[-1].SetAlpha(0.0)
filter_list[-1].SetBeta(1.0)
aug_image_lists = [] # Used only for display purposes in this notebook.
for i, img in enumerate(image_list):
aug_image_lists.append([f.Execute(img) for f in filter_list])
for aug_image, f in zip(aug_image_lists[-1], filter_list):
sitk.WriteImage(
aug_image, output_prefix + str(i) + '_' + f.GetName() + '.' +
output_suffix)
return aug_image_lists
def data_augumentation_noise(img_3d):
# Create a list of intensity modifying filters, which we apply to the given images
filter_list = []
# Smoothing filters
filter_list.append(sitk.SmoothingRecursiveGaussianImageFilter())
filter_list[-1].SetSigma(2.0)
filter_list.append(sitk.DiscreteGaussianImageFilter())
filter_list[-1].SetVariance(4.0)
filter_list.append(sitk.BilateralImageFilter())
filter_list[-1].SetDomainSigma(4.0)
filter_list[-1].SetRangeSigma(8.0)
filter_list.append(sitk.MedianImageFilter())
filter_list[-1].SetRadius(8)
# Noise filters using default settings
# Filter control via SetMean, SetStandardDeviation.
filter_list.append(sitk.AdditiveGaussianNoiseImageFilter())
filter_list[-1].SetMean(0.2)
filter_list[-1].SetStandardDeviation(0.7)
# Filter control via SetProbability
filter_list.append(sitk.SaltAndPepperNoiseImageFilter())
# Filter control via SetScale
filter_list.append(sitk.ShotNoiseImageFilter())
# Filter control via SetStandardDeviation
filter_list.append(sitk.SpeckleNoiseImageFilter())
filter_list.append(sitk.AdaptiveHistogramEqualizationImageFilter())
filter_list[-1].SetAlpha(1.0)
filter_list[-1].SetBeta(0.0)
filter_list.append(sitk.AdaptiveHistogramEqualizationImageFilter())
filter_list[-1].SetAlpha(0.0)
filter_list[-1].SetBeta(1.0)
aug_image_lists = [] # Used only for display purposes in this notebook.
f = filter_list[4]
aug_image = f.Execute(img_3d)
return aug_image
def histogram_equalize(image, radius=(1, 1, 1), alpha=0.6, beta=0.3):
if isinstance(image, (np.ndarray)):
image = sitk.GetImageFromArray(image)
filter = sitk.AdaptiveHistogramEqualizationImageFilter()
heq = sitk.AdaptiveHistogramEqualization(image,
radius,
alpha=alpha,
beta=beta)
# mean = sitk.Mean(heq,(1,1))
return sitk.GetArrayFromImage(heq)
def aeh(image, tup=tuple):
#Standard Implementation: aeh(image,0.9,0.6.255,0)
adapt = sitk.AdaptiveHistogramEqualizationImageFilter()
adapt.SetAlpha(tup[0])
adapt.SetBeta(tup[1])
image = adapt.Execute(image) # set mean and std deviation
normalizeFilter = sitk.NormalizeImageFilter()
resacleFilter = sitk.RescaleIntensityImageFilter()
resacleFilter.SetOutputMaximum(tup[2])
resacleFilter.SetOutputMinimum(tup[3])
image = resacleFilter.Execute(image) # set intensity 0-255
image = normalizeFilter.Execute(image) # set mean and std deviation
return image
def __call__(self, sample):
adapt = sitk.AdaptiveHistogramEqualizationImageFilter()
adapt.SetAlpha(0.7)
adapt.SetBeta(0.8)
image = adapt.Execute(sample['image']) # set mean and std deviation
resacleFilter = sitk.RescaleIntensityImageFilter()
resacleFilter.SetOutputMaximum(255)
resacleFilter.SetOutputMinimum(0)
image = resacleFilter.Execute(image) # set mean and std deviation
label = sample['label']
return {'image': image, 'label': label}
def adapt_eq_histogram(image):
adapt = sitk.AdaptiveHistogramEqualizationImageFilter()
adapt.SetAlpha(0.9)
adapt.SetBeta(0.6)
image = adapt.Execute(image) # set mean and std deviation
normalizeFilter = sitk.NormalizeImageFilter()
resacleFilter = sitk.RescaleIntensityImageFilter()
resacleFilter.SetOutputMaximum(255)
resacleFilter.SetOutputMinimum(0)
image = normalizeFilter.Execute(image) # set mean and std deviation
image = resacleFilter.Execute(image) # set intensity 0-255
return image
def aeh(image,alpha,beta,outMax,outMin):
#Standard Implementation: aeh(image,0.9,0.6.255,0)
adapt = sitk.AdaptiveHistogramEqualizationImageFilter()
adapt.SetAlpha(alpha)
adapt.SetBeta(beta)
image = adapt.Execute(image) # set mean and std deviation
normalizeFilter = sitk.NormalizeImageFilter()
resacleFilter = sitk.RescaleIntensityImageFilter()
resacleFilter.SetOutputMaximum(255)
resacleFilter.SetOutputMinimum(0)
image = normalizeFilter.Execute(image) # set mean and std deviation
image = resacleFilter.Execute(image) # set intensity 0-255
return image
arrCleaned = arr1 * arr2
return sitk.GetImageFromArray(arrCleaned)
def findBladderCenter(image, edgeImage, borderImage):
image = sitk.GetArrayFromImage(image)
edgeImage = sitk.GetArrayFromImage(edgeImage)
borderImage = sitk.GetArrayFromImage(borderImage) == 0
#since borderImage tLarge has the inside of the area marked 1, the outside 0, we take the negative
out = image + edgeImage
out = out + (borderImage * 255)
return sitk.GetImageFromArray(out)
#%%
aehFilter = sitk.AdaptiveHistogramEqualizationImageFilter()
aehFilter.SetAlpha(1.3)
aehFilter.SetBeta(1)
aehFilter.SetRadius(5)
aehImg = aehFilter.Execute(imgResult)
arrAeh = sitk.GetArrayFromImage(aehImg)
arrAehCleaned = arrAeh * arrExcl
imgAehCleaned = sitk.GetImageFromArray(arrAehCleaned)
#plotSlices(imgAehCleaned,'sagittal')
#%%
gaussian.SetSigma(3)
imOut = gaussian.Execute(imgAehCleaned)
#plotSlices(imOut,'sagittal')
#%%Array Magic
statF2 = sitk.StatisticsImageFilter()
