def otsu_threshold(im, maxCount=255, verbose=False):
"""
Function to turn an image into a binary split based on otus's method
Args:
im (array): image to threshold
maxCount (int): maximum value of any element in the array
For images this will be 255
verbose (boolean): flag to visually show plot of histogram
with threshold
Returns:
a tuple of a binary image (array), and the threshold (int)
Raises:
TypeError: if image is not a numpy ndarray
ValueError: if image is not in the shape of a grayscale image
"""
# type checking, look at the above Raises section
if (not isinstance(im, np.ndarray)):
raise TypeError("image is not a numpy ndarray; use openCV's imread")
if (len(np.shape(im)) != 2):
raise ValueError(
"Shape of image is not a grayscale image or histogram")
# get last non-zero pixel level
hist = cv2.calcHist([im], [0], None, [255], [0, 255])
# get the element pair, which contains an index and value (which is 0)
firstPixelIndex = np.transpose(np.nonzero(hist))[0][0]
lastPixelIndex = np.transpose(np.nonzero(hist))[-1][0]
# evaluate sigma of b squared for all values from the first non zero pixel
# level to the last non zero pixel level
result = np.zeros(maxCount)
for i in range(firstPixelIndex + 1, lastPixelIndex):
sigma = class_variance_b_squared(im, i, maxCount)
result[i] = sigma
ipcv.plotHist(result)
thresh = np.argmax(result)
if (verbose):
ipcv.plotHist(hist, [np.argmax(result)])
# apply threshold to image
binaryIm = np.ones(np.shape(im))
binaryIm = binaryIm * im
np.place(binaryIm, im < thresh, 0)
np.place(binaryIm, im >= thresh, 1)
return (binaryIm, thresh)
def otsu_threshold(im, maxCount=255, verbose=False):
"""
Function to turn an image into a binary split based on otus's method
Args:
im (array): image to threshold
maxCount (int): maximum value of any element in the array
For images this will be 255
verbose (boolean): flag to visually show plot of histogram
with threshold
Returns:
a tuple of a binary image (array), and the threshold (int)
Raises:
TypeError: if image is not a numpy ndarray
ValueError: if image is not in the shape of a grayscale image
"""
# type checking, look at the above Raises section
if (not isinstance(im, np.ndarray)):
raise TypeError("image is not a numpy ndarray; use openCV's imread")
if (len(np.shape(im)) != 2):
raise ValueError("Shape of image is not a grayscale image or histogram")
# get last non-zero pixel level
hist = cv2.calcHist([im],[0],None,[255],[0,255])
# get the element pair, which contains an index and value (which is 0)
firstPixelIndex = np.transpose(np.nonzero(hist))[0][0]
lastPixelIndex = np.transpose(np.nonzero(hist))[-1][0]
# evaluate sigma of b squared for all values from the first non zero pixel
# level to the last non zero pixel level
result = np.zeros(maxCount)
for i in range(firstPixelIndex+1, lastPixelIndex):
sigma = class_variance_b_squared(im, i, maxCount)
result[i] = sigma
ipcv.plotHist(result)
thresh = np.argmax(result)
if (verbose):
ipcv.plotHist(hist, [np.argmax(result)])
# apply threshold to image
binaryIm = np.ones(np.shape(im))
binaryIm = binaryIm*im
np.place(binaryIm, im<thresh, 0)
np.place(binaryIm, im>=thresh, 1)
return (binaryIm, thresh)
def u(im, k):
"""
Function to evaulate the class mean levels
Args:
im (array): image to convert to histogram and evaluate
k (int): pixel level to evaluating mean levels
Returns:
class mean level for k
"""
# get histogram of probabilities
hist = cv2.calcHist([im], [0], None, [255 + 1], [0, 255 + 1])
histProbs = hist / np.cumprod(np.shape(im))[-1]
# build array of indicies
indHist = np.arange(256)
histMean = indHist * histProbs
ipcv.plotHist(histProbs)
ipcv.plotHist(indHist)
ipcv.plotHist(histMean)
result = np.cumsum(histMean)[k]
return result
def u(im, k):
"""
Function to evaulate the class mean levels
Args:
im (array): image to convert to histogram and evaluate
k (int): pixel level to evaluating mean levels
Returns:
class mean level for k
"""
# get histogram of probabilities
hist = cv2.calcHist([im],[0],None,[255+1],[0,255+1])
histProbs = hist / np.cumprod(np.shape(im))[-1]
# build array of indicies
indHist = np.arange(256)
histMean = indHist * histProbs
ipcv.plotHist(histProbs)
ipcv.plotHist(indHist)
ipcv.plotHist(histMean)
result = np.cumsum(histMean)[k]
return result