def compositeThreshold(gray, mode='com'):
if mode == 'otsu':
otsu = threshold_otsu(gray)
otsu_bin = gray > otsu
otsu_bin = otsu_bin.astype(np.uint8) * 255
return otsu_bin
elif mode == 'yen':
yen = threshold_yen(gray)
yen_bin = gray > yen
yen_bin = yen_bin.astype(np.uint8) * 255
return yen_bin
elif mode == 'li':
li = threshold_li(gray)
li_bin = gray > li
li_bin = li_bin.astype(np.uint8) * 255
return li_bin
elif mode == 'niblack':
niblack = threshold_niblack(gray, window_size=13, k=0.8)
niblack_bin = gray > niblack
niblack_bin = niblack_bin.astype(np.uint8) * 255
return niblack_bin
elif mode == 'sauvola':
sauvola = threshold_sauvola(gray, window_size=13)
sauvola_bin = gray > sauvola
sauvola_bin = sauvola_bin.astype(np.uint8) * 255
return sauvola_bin
elif mode == 'com':
li = threshold_li(gray)
li_bin = gray > li
li_bin = li_bin.astype(np.uint8) * 255
otsu = threshold_otsu(gray)
otsu_bin = gray > otsu
otsu_bin = otsu_bin.astype(np.uint8) * 255
yen = threshold_yen(gray)
yen_bin = gray > yen
yen_bin = yen_bin.astype(np.uint8) * 255
return cv2.min(cv2.min(otsu_bin, li_bin), yen_bin)
elif mode == "niblack-multi":
thr = np.zeros((gray.shape), dtype=np.uint8)
thr[thr >= 0] = 255
for k in np.linspace(-0.8, 0.2, 5): #(-1.8,0.2,5)
thresh_niblack = threshold_niblack(gray, window_size=25, k=k)
binary_niblack = gray > thresh_niblack
binary_niblack = binary_niblack.astype(np.uint8) * 255
showResult("binary_niblack", binary_niblack)
thr = cv2.min(thr, binary_niblack)
return thr
else:
sauvola = threshold_sauvola(gray, window_size=25, k=0.25)
sauvola_bin = gray > sauvola
sauvola_bin = sauvola_bin.astype(np.uint8) * 255
niblack = threshold_niblack(gray, window_size=25, k=0.25)
niblack_bin = gray > niblack
niblack_bin = niblack_bin.astype(np.uint8) * 255
return cv2.max(sauvola, niblack)
def SauvolaModBinarization(image, n1=51, n2=51, k1=0.3, k2=0.3, default=True):
'''
Binarization using Sauvola's algorithm
@name : SauvolaModBinarization
parameters
@param image (numpy array of shape (3/1) of type np.uint8): color or gray scale image
optional parameters
@param n1 (int) : window size for running sauvola during the first pass
@param n2 (int): window size for running sauvola during the second pass
@param k1 (float): k value corresponding to sauvola during the first pass
@param k2 (float): k value corresponding to sauvola during the second pass
@param default (bool) : bollean variable to set the above parameter as default.
@param default is set to True : thus default values of the above optional parameters (n1,n2,k1,k2) are set to
n1 = 5 % of min(image height, image width)
n2 = 10 % of min(image height, image width)
k1 = 0.5
k2 = 0.5
Returns
@return A binary image of same size as @param image
@cite https://drive.google.com/file/d/1D3CyI5vtodPJeZaD2UV5wdcaIMtkBbdZ/view?usp=sharing
'''
if (default):
n1 = int(0.05 * min(image.shape[0], image.shape[1]))
if (n1 % 2 == 0):
n1 = n1 + 1
n2 = int(0.1 * min(image.shape[0], image.shape[1]))
if (n2 % 2 == 0):
n2 = n2 + 1
k1 = 0.5
k2 = 0.5
if (image.ndim == 3):
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
else:
gray = np.copy(image)
T1 = threshold_sauvola(gray, window_size=n1, k=k1)
max_val = np.amax(gray)
min_val = np.amin(gray)
C = np.copy(T1)
C = C.astype(np.float32)
C[gray > T1] = (gray[gray > T1] - T1[gray > T1]) / (max_val -
T1[gray > T1])
C[gray <= T1] = 0
C = C * 255.0
new_in = np.copy(C.astype(np.uint8))
T2 = threshold_sauvola(new_in, window_size=n2, k=k2)
binary = np.copy(gray)
binary[new_in <= T2] = 0
binary[new_in > T2] = 255
return binary
def sauvola(grayimg, w=51, k=0.2, scaledown=None, reverse=False):
mask =None
if scaledown is not None:
mask = cv2.resize(grayimg,None,fx=scaledown,fy=scaledown)
w = int(w * scaledown)
if w % 2 == 0: w += 1
mask = threshold_sauvola(mask, w, k)
mask = cv2.resize(mask,(grayimg.shape[1],grayimg.shape[0]),fx=scaledown,fy=scaledown)
else:
if w % 2 == 0: w += 1
mask = threshold_sauvola(grayimg, w, k)
if reverse:
return where(grayimg > mask, uint8(0), uint8(1))
else:
return where(grayimg > mask, uint8(1), uint8(0))
def processBinarization(self, algorithm=ImageAlgorithm.SAUVOLA):
if algorithm == ImageAlgorithm.SAUVOLA:
image = skimage.io.imread(fname=self.pathImg, as_gray=True)
thresh_sauvola = threshold_sauvola(image, window_size=81)
self.binary_sauvola = image > thresh_sauvola
"""
self.binary_sauvola = invert(self.binary_sauvola)
chull = convex_hull_image(self.binary_sauvola)
[rows, columns] = np.where(chull)
EPS = 50
row1 = min(rows) - EPS
row2 = max(rows) + EPS
col1 = min(columns) - EPS
col2 = max(columns) + EPS
self.binary_sauvola = self.binary_sauvola[row1:row2, col1:col2]
self.binary_sauvola = invert(self.binary_sauvola)
"""
self.binary_sauvola = invert(self.binary_sauvola)
#selem = disk(6)
self.binary_sauvola = thin(self.binary_sauvola, np.int(15))
self.binary_sauvola = np.invert(self.binary_sauvola)
#thresh_sauvola = threshold_sauvola(self.binary_sauvola, window_size=21)
#self.binary_sauvola = self.binary_sauvola > thresh_sauvola
self.binary_sauvola = erosion(self.binary_sauvola)
#self.binary_sauvola = gaussian(self.binary_sauvola)
elif algorithm == ImageAlgorithm.OTSU:
self.otsuBinarization()
def get_leaf_venation(gray_img):
# using the sauvola algorithm
threshold = threshold_sauvola(image=gray_img, window_size=11, k=0.04)
venation_img = gray_img > threshold
plt.imshow(venation_img)
plt.show()
return venation_img
def binarize(image: np.ndarray, holes_threshold: float = 20) -> np.ndarray:
"""
Binarize image using Sauvola algorithm.
Parameters
----------
image : np.ndarray
RGB image to binarize.
holes_threshold : float, optional
Pixel areas covering less than the given number of pixels are removed in the process, by default 20.
Returns
-------
binarized : np.ndarray
Binarized and filtered image.
"""
# Extract brightness channel from HSV-converted image
image_gray = rgb2hsv(image)[:,:,2]
# Enhance contrast
image_gray = equalize_adapthist(image_gray, kernel_size=100)
# Threshold using Sauvola algorithm
thresh_sauvola = threshold_sauvola(image_gray, window_size=51, k=0.25)
binary_sauvola = image_gray > thresh_sauvola
# Remove small objects
binary_cleaned = 1.0 * remove_small_holes(binary_sauvola, area_threshold=holes_threshold)
# Remove thick black border (introduced during thresholding)
binary_cleaned = flood_fill(binary_cleaned, (0, 0), 0)
binary_cleaned = flood_fill(binary_cleaned, (0, 0), 1)
return binary_cleaned.astype(np.bool)
def binarise_sauvola(image, window_size = 59, k = 0.5, r = 128):
thresh_sauvola = threshold_sauvola(image, window_size=window_size, k=k, r=r)
binarised_sauvola = image > thresh_sauvola
binarised_sauvola = img_as_ubyte(binarised_sauvola)
return binarised_sauvola
def fit(self, X=None, y=None):
if self.threshold_type == "otsu":
threshold_function = lambda data: filters.threshold_otsu(data)
elif self.threshold_type == "local_otsu":
threshold_function = lambda data: filters.rank.otsu(data, morphology.square(self.block_size))
elif self.threshold_type == "local":
threshold_function = lambda data: filters.threshold_local(data, self.block_size)
elif self.threshold_type == "niblack":
threshold_function = lambda data: filters.threshold_niblack(data, self.block_size)
elif self.threshold_type == "sauvola":
threshold_function = lambda data: filters.threshold_sauvola(data, self.block_size)
elif self.threshold_type is None:
threshold_function = None
else:
raise ValueError("Unknown threshold type: {}".format(self.threshold_type))
self.ocr_reader_ = screen_ocr.Reader.create_reader(
backend=self.backend,
threshold_function=threshold_function,
correction_block_size=self.correction_block_size,
margin=self.margin,
resize_factor=self.resize_factor,
resize_method=self.resize_method,
convert_grayscale=self.convert_grayscale,
shift_channels=self.shift_channels,
label_components=self.label_components,
debug_image_callback=None)
def main(img, ws):
image = img
binary_global = image > threshold_otsu(image)
window_size = ws
thresh_niblack = threshold_niblack(image, window_size=window_size, k=0.8)
thresh_sauvola = threshold_sauvola(image, window_size=window_size)
binary_niblack = image > thresh_niblack
binary_sauvola = image > thresh_sauvola
plt.figure(figsize=(8, 7))
plt.subplot(2, 2, 1)
plt.imshow(image, cmap=plt.cm.gray)
plt.title('Original')
plt.axis('off')
plt.subplot(2, 2, 2)
plt.title('Global Threshold')
plt.imshow(binary_global, cmap=plt.cm.gray)
plt.axis('off')
plt.subplot(2, 2, 3)
plt.imshow(binary_niblack, cmap=plt.cm.gray)
plt.title('Niblack Threshold')
plt.axis('off')
plt.subplot(2, 2, 4)
plt.imshow(binary_sauvola, cmap=plt.cm.gray)
plt.title('Sauvola Threshold')
plt.axis('off')
return plt, binary_global, binary_niblack, binary_sauvola
def threshold(image,
*,
sigma=0.,
radius=0,
offset=0.,
method='sauvola',
smooth_method='Gaussian'):
"""Use scikit-image filters to "intelligently" threshold an image.
Parameters
----------
image : array, shape (M, N, ...[, 3])
Input image, conformant with scikit-image data type
specification [1]_.
sigma : float, optional
If positive, use Gaussian filtering to smooth the image before
thresholding.
radius : int, optional
If given, use local median thresholding instead of global.
offset : float, optional
If given, reduce the threshold by this amount. Higher values
result in fewer pixels above the threshold.
method: {'sauvola', 'niblack', 'median'}
Which method to use for thresholding. Sauvola is 100x faster, but
median might be more accurate.
smooth_method: {'Gaussian', 'TV'}
Which method to use for smoothing. Choose from Gaussian smoothing
and total variation denoising.
Returns
-------
thresholded : image of bool, same shape as `image`
The thresholded image.
References
----------
.. [1] http://scikit-image.org/docs/dev/user_guide/data_types.html
"""
if sigma > 0:
if smooth_method.lower() == 'gaussian':
image = filters.gaussian(image, sigma=sigma)
elif smooth_method.lower() == 'tv':
image = restoration.denoise_tv_bregman(image, weight=sigma)
if radius == 0:
t = filters.threshold_otsu(image) + offset
else:
if method == 'median':
footprint = hyperball(image.ndim, radius=radius)
t = ndi.median_filter(image, footprint=footprint) + offset
elif method == 'sauvola':
w = 2 * radius + 1
t = threshold_sauvola(image, window_size=w, k=offset)
elif method == 'niblack':
w = 2 * radius + 1
t = threshold_niblack(image, window_size=w, k=offset)
else:
raise ValueError('Unknown method %s. Valid methods are median,'
'niblack, and sauvola.' % method)
thresholded = image > t
return thresholded
def threshold(img):
cpy = img.copy()
blk = max(cpy.shape[0], cpy.shape[1])
blk = int(blk * 0.03) + 1 if int(blk * 0.03) % 2 == 0 else int(blk * 0.03)
thresh_sauvola = threshold_sauvola(cpy, blk)
binary_sauvola = cpy > thresh_sauvola
return binary_sauvola
def main():
image = cv2.imread("./counter_images/01305.png", cv2.IMREAD_GRAYSCALE)
# image = cv2.imread("../venv/lib/python3.7/site-packages/skimage/data/page.png", cv2.IMREAD_GRAYSCALE)
image = 255 - image
image = cv2.medianBlur(image, 7)
binary_global = image > threshold_otsu(image)
binary_global = np.uint8(binary_global * 255)
window_size = 41
thresh_niblack = threshold_niblack(image, window_size=window_size, k=0.8)
thresh_sauvola = threshold_sauvola(image, window_size=window_size, k=.1)
binary_niblack = image > thresh_niblack
binary_niblack = np.uint8(binary_niblack * 255)
binary_sauvola = image > thresh_sauvola
binary_sauvola = np.uint8(binary_sauvola * 255)
imshowWait(image=image,
binary_global=binary_global,
binary_niblack=binary_niblack,
binary_sauvola=binary_sauvola)
def transform_img(img, img_rows, img_cols, mnist=False):
# Binarize
thresh = threshold_sauvola(img, window_size=13, k=0.025, r=0.5)
img = (img < thresh)
[h, w] = img.shape
# Pad
if w > h:
factor = img_cols / w
img = rescale(img, factor, mode='constant', cval=0)
diff = (img_rows - img.shape[0]) / 2
img = np.pad(img, ((int(ceil(diff)), int(floor(diff))), (0, 0)),
'constant',
constant_values=((0, )))
else:
factor = img_rows / h
img = rescale(img, factor, mode='constant', cval=0)
diff = (img_cols - img.shape[1]) / 2
img = np.pad(img, ((0, 0), (int(ceil(diff)), int(floor(diff)))),
'constant',
constant_values=((0, )))
# Binarize again
if mnist:
ret = (img > 0.4).astype(int)
else:
ret = binary_dilation(img)
ret = ret.astype(int)
return ret
def preprocessImage(img_path):
# preprocess and load input image
img = cv2.imread(img_path)
resized_image = cv2.resize(img, (250, 250))
# removing noise form colored images
denoised_image = cv2.fastNlMeansDenoisingColored(resized_image, None, 10,
10, 7, 21)
denoised_img_name = imgName()
cv2.imwrite("online_data_collection/image_processing/" + denoised_img_name,
denoised_image)
# Since denoiseing converts BGR image to CELAB, converting CELAB to BGR
bgr_image = cv2.cvtColor(denoised_image, cv2.COLOR_Lab2BGR)
bgr_img_name = imgName()
cv2.imwrite("online_data_collection/image_processing/" + bgr_img_name,
bgr_image)
# Converion of BGR to Grayscale
gray_image = cv2.cvtColor(bgr_image, cv2.COLOR_BGR2GRAY)
gray_img_name = imgName()
cv2.imwrite("online_data_collection/image_processing/" + gray_img_name,
gray_image)
# Binarization
thres_sauvola = threshold_sauvola(gray_image, window_size=17)
binary_sauvola = gray_image > thres_sauvola
binary_sauvola = img_as_ubyte(binary_sauvola)
binary_img_name = imgName()
cv2.imwrite("online_data_collection/image_processing/" + binary_img_name,
binary_sauvola)
# Smoothing image
# smooth_image = cv2.medianBlur(binary_sauvola, 3)
smooth_image = cv2.bilateralFilter(binary_sauvola, 9, 75, 75)
smooth_img_name = imgName()
cv2.imwrite("online_data_collection/image_processing/" + smooth_img_name,
smooth_image)
return denoised_img_name, bgr_img_name, gray_img_name, binary_img_name, smooth_img_name
def preprocess_image(img, pipeline_id="baseline"):
if pipeline_id == "direct_bit_extract":
# No preprocessing wanted
return img
# dpi = img.info["dpi"]
# Edge enhance
img = img.filter(ImageFilter.EDGE_ENHANCE_MORE)
# Thresholding
img = skimage.img_as_ubyte(img)
# Tests
# from skimage.filters import (try_all_threshold, threshold_sauvola,
# threshold_local, threshold_niblack)
# import matplotlib.pyplot as plt
# tmp = skimage.img_as_ubyte(img > threshold_local(img, 15))
# Image.fromarray(tmp).show()
# tmp = skimage.img_as_ubyte(img > threshold_niblack(img))
# Image.fromarray(tmp).show()
# tmp = skimage.img_as_ubyte(img > threshold_sauvola(img))
# Image.fromarray(tmp).show()
# fig, ax = try_all_threshold(img, verbose=False)
# plt.show()
img = skimage.img_as_ubyte(img > threshold_sauvola(img))
img = Image.fromarray(img)
# img.info["dpi"] = dpi
return img
def thresh(image,threshtype='local', make_binary=True, simple=False):
im=image.copy()
if simple==True:
im[im<0]=0
return im
if threshtype=='global':
from skimage.filters import threshold_otsu
try:
thresh = threshold_otsu(im)
except ValueError:
thresh=0
binary= im>thresh
if make_binary==False:
return binary
return binary.astype(int)
if threshtype=='local':
from skimage.filters import threshold_sauvola
thresh = threshold_sauvola(im)
binary = im< thresh
if make_binary==False:
return binary
return binary.astype(int)
def binarize_sauvola(image):
#binarized_filename = input_image_filename.split(".")[0] + "_bin.png"
# image = page()
# image binarized_filename
# image = io.imread(input_image_filename)
# opening and grayscaling the image
#image = cv2.imread(input_image_filename, cv2.IMREAD_GRAYSCALE)
#thresh_niblack = threshold_niblack(image, k=0.8)
thresh_sauvola = threshold_sauvola(image)
#binary_niblack = image > thresh_niblack
binary_sauvola = image > thresh_sauvola
#binary_niblack = binary_niblack.astype(int)
binary_sauvola = binary_sauvola.astype(int)
#binary_niblack *= 255
binary_sauvola *= 255
# applying the binarization
# image_bin = cv2.threshold(image, 0, 255, cv2.THRESH_BINARY + cv2.THRESH_OTSU)[1]
#cv2.imwrite("niblack.png", binary_niblack)
cv2.imwrite("sauvola.png", binary_sauvola)
#print("Binarization completed! Binarized image saved into ", binarized_filename)
#return binarized_filename
def __sauvola_method(image: Union[np.ndarray, Iterable, np.uint8], stream,
progres) -> np.ndarray:
th = threshold_sauvola(image, window_size=45)
im = image <= th
# imshow(im, cmap='gray')
# plt.show()
progres.progress(33)
# stream[1].append(im)
# stream[0].image(stream[1], width=300)
# if im.shape[0] > 1000:
# gaus = 7
# elif im.shape[0] > 650:
# gaus = 3
# else:
# gaus = 1
if im.shape[0] > 400:
gaus = int(0.00330957 * im.shape[0] - 0.13687352)
else:
gaus = 1
result = gaussian(im, gaus)
# imshow(result)
# plt.show()
progres.progress(66)
stream[1].append(result)
stream[0].image(stream[1], width=300)
# result = minimum(maximum(result, disk(5)), disk(12))
# imshow(result)
# plt.show()
return result
def blob_detection2(post_masked_processed, draw):
""" Funtion that performs blob detection to localize electrodes. Does not include dilation and erosion. For use in problematic images where blobs are hard to locate.
Inputs : Processed post oeprative image, draw (Bool)
Outputs : If draw is True, outputs an image with the blobs drawn
If draw is False, outputs the coordinates of the centers of the blobs
"""
# Initial set up; no erosion or dilationn
img = cv2.normalize(post_masked_processed, None, 0, 255, cv2.NORM_MINMAX)
img = img.astype('uint8')
img = cv2.medianBlur(img, 11)
th2 = filters.threshold_sauvola(img)
th2 = 255 - th2
th2 = th2.astype("uint8")
# Set our filtering parameters
# Initialize parameter settiing using cv2.SimpleBlobDetector
params = cv2.SimpleBlobDetector_Params()
# Set Area filtering parameters
params.filterByArea = True
params.minArea = 10
# Set Circularity filtering parameters
params.filterByCircularity = False
params.minCircularity = 0.1
# Set Convexity filtering parameters
params.filterByConvexity = True
params.minConvexity = 0.1
# Set inertia filtering parameters
params.filterByInertia = True
params.minInertiaRatio = 0.01
# Create a detector with the parameters
detector = cv2.SimpleBlobDetector_create(params)
# Detect blobs
keypoints = detector.detect(th2)
if draw == True:
# Draw blobs on our image as red circles
blank = np.zeros((1, 1))
blobs = cv2.drawKeypoints(th2, keypoints, blank, (0, 0, 255),
cv2.DRAW_MATCHES_FLAGS_DRAW_RICH_KEYPOINTS)
number_of_blobs = len(keypoints)
text = "Number of Circular Blobs: " + str(len(keypoints))
cv2.putText(blobs, text, (20, 550), cv2.FONT_HERSHEY_SIMPLEX, 1,
(0, 100, 255), 2)
return blobs
else:
result = []
for point in keypoints:
x = point.pt[0]
y = point.pt[1]
result.append([x, y])
return result
def ThresholdSauvola(img):
## based on observation using these values of k,r work best (default k=0.2)
thresh_sauvola = threshold_sauvola(img.copy(), window_size=31, k=0.1, r=45)
binary_sauvola = img > thresh_sauvola
image = binary_sauvola.astype('uint8')*255
## Sauvola is edge sensitive and prone to creating thick borders around image. Try remove them based on size
contours, _ = cv2.findContours(255-image.copy(),cv2.RETR_LIST,cv2.CHAIN_APPROX_NONE)
mask = np.ones(img.shape[:2], dtype="uint8") * 255
mask_remove = cv2.imread('image.jpg', cv2.IMREAD_ANYCOLOR)
mask_retain = cv2.imread('image.jpg', cv2.IMREAD_ANYCOLOR)
# loop over the contours
for c in contours:
# if the contour is bad, draw it on the mask
if is_good_contour_for_line_segmentation(c, 255-image.copy())==True:
r = cv2.boundingRect(c)
cv2.rectangle(mask, (r[0],r[1]), (r[0]+r[2],r[1]+r[3]), color=0, thickness=-1)
cv2.drawContours(mask_retain, [c], 0, color=(0,255,0), thickness=1)
else:
#r = cv2.boundingRect(c)
cv2.drawContours(mask_remove, [c], 0, color=(0,0,255), thickness=-1)
showimage(mask_retain, 'retaining-this')
showimage(mask_remove, 'removing-this')
# remove the contours from the image and show the resulting images
showimage(255-mask,'contour-removal-mask')
## AARGH! someone please simplify this! X is the correct form, it has already been inverted so text is white.
x = cv2.bitwise_and(255-image.copy(),255-mask, mask=255-mask)
return x
def process_regions(image, blur_sigma=3, opening_size=3, orientation_deviation=15, overlap_minimum=0.8):
'''
Attempt to find any possible marker corner regions in a given image
Inputs:
- image: grayscale image that may contain a marker
- blur_sigma: parameter for Gaussian blur to use on image
- opening_size: parameter for morphological opening to use on image
- orientation_deviation: see orientation parameter used by region_filter_heuristic(...)
- overlap_minimum: see similarity parameter used by region_filter_heuristic(...)
Returns: a 2-tuple of:
- the image after pre-processing steps like blurring, thresholding, etc.
- the list of regionprops that may be possible marker corners
'''
# Blur and equalize the image
image = exposure.equalize_hist(image)
image = filters.gaussian(image, sigma=blur_sigma)
# Use local thresholding
image = (image <= filters.threshold_sauvola(image, k=0.1))
image = morphology.opening(image, selem=morphology.disk(opening_size))
# Label components in the image
labeled = measure.label(image, connectivity=2)
components = measure.regionprops(labeled, intensity_image=image)
# Sort the components by our rectangle heuristic
return image, labeled, [r for r in components if region_filter_heuristic(r, orientation_deviation, overlap_minimum)]
def thickness_score(im, window_size=5, threshold=10, name=None):
if im.ndim == 3:
im = cv2.cvtColor(im.astype(np.uint8), cv2.COLOR_BGR2GRAY)
cv2.imwrite('raw_' + name, im)
edge = cv2.Canny(im,50,100)
cv2.imwrite('edge_' + name, edge)
e_pixel_count = 0
for i in range(edge.shape[0]):
for j in range(edge.shape[1]):
if edge[i][j] == 255:
e_pixel_count += 1
# 计算感兴趣的区域（文字区域）
# threshold_val, _ = cv2.threshold(im, -1, 255, cv2.THRESH_OTSU)
threshold_val = threshold_sauvola(im, window_size=15, k=0.15)
roi = np.zeros_like(im)
roi = np.array(im < threshold_val + 10).astype(np.int32) # roi范围扩大一些？
roi_pixel_count = 0
for i in range(roi.shape[0]):
for j in range(roi.shape[1]):
if roi[i][j] == 1:
roi_pixel_count += 1
cv2.imwrite('roi_' + name, roi*255)
stroke_width = 0
if e_pixel_count == 0:
stroke_width = -1
else:
stroke_width = float(roi_pixel_count / e_pixel_count)
print('stroke width:', stroke_width)
return stroke_width
def cod16(image):
# hint: increase contrast.
window_size = 3
# image = cv.imread("sample2.jpg")
# image = cv.cvtColor(image, cv.COLOR_RGB2GRAY)
# ix = lambda i: cv.cvtColor(i, cv.COLOR_RGB2GRAY)
# print(type(image),image.shape)
# print(image)
image = clahe_demo(image)
# must use gray.
# print("spliter")
# that is f*****g insane.
thresh_niblack = (
toReal(threshold_niblack(image, window_size=window_size, k=3)) *
255.0).astype(np.uint8)
thresh_sauvola0 = (
toReal(threshold_sauvola(image, window_size=3, k=1.303)) *
255.0).astype(np.uint8) # for full scan?
# perfect inverse of the one above
thresh_sauvola1 = (toReal(
threshold_sauvola(image, window_size=3, k=0.999999999999999999)) *
255.0).astype(np.uint8) # for button detection?
# just filter those things out.
# thresh_niblack = threshold_niblack(image, window_size=window_size, k=12).astype(np.uint8)
# thresh_sauvola0 = threshold_sauvola(image, window_size=3, k=1.103).astype(np.uint8) # for full scan?
# # perfect inverse of the one above
# thresh_sauvola1 = threshold_sauvola(image, window_size=3, k=1.523).astype(np.uint8) # for button detection?
# print(type(thresh_niblack), thresh_niblack.shape)
# print(type(thresh_niblack[0][0]))
# ndarray.
# it just cannot be right. give it up?
# the shape can differ.
# calcMe(thresh_niblack)
# calcMe(thresh_sauvola0)
# calcMe(thresh_sauvola1)
# cv.imshow("grayimage",thresh_niblack)
# cv.waitKey(0)
# cv.imshow("thresholdimage", thresh_sauvola0)
# cv.waitKey(0)
# cv.imshow("thresholdimage", thresh_sauvola1)
# cv.waitKey(0)
# cv.destroyAllWindows()
# return
# # use floor function?
# # just what the f**k?
return thresh_niblack, thresh_sauvola0, thresh_sauvola1
def doThresholding(method, window_size, k, img):
if method == 'niblack':
th_mask = threshold_niblack(img, window_size=window_size, k=k)
if method == 'sauvola':
th_mask = threshold_sauvola(img, window_size=window_size)
binary_mask = img < th_mask
return(binary_mask)
def binarize(img, window_size=35):
# cv_image = img_as_ubyte(img)
# img = cv2.adaptiveThreshold(cv_image, 255, cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY, window_size, 0)
thresh = filters.threshold_sauvola(img, window_size)
img = img > thresh
img = np.array(img * 255, dtype=np.uint8)
return img
def sauvola(img, window_size=25, min_threshold=0.02):
k = 0.25
r = 0.5
t = threshold_sauvola(img, window_size=window_size, k=k, r=r)
if min_threshold is not None:
t[t <= min_threshold] = min_threshold
return t
def extract_feature(arr):
radius = 1
n_points = radius * 8
arr = cv2.cvtColor(arr, cv2.COLOR_BGR2GRAY)
distances = [1, 5]
angles = [0, np.pi / 4, np.pi / 2, 3 * np.pi / 4]
glcm = greycomatrix(arr,
distances=distances,
angles=angles,
levels=256,
symmetric=False,
normed=False)
# properties = ['dissimilarity', 'homogeneity', 'contrast', 'ASM', 'energy', 'correlation']
# glcm_feats = np.hstack([greycoprops(glcm, prop=prop).ravel() for prop in properties])
glcm_feats = np.hstack(
[adadoc.greycoprops(glcm[:, :, i, :]) for i in range(0, 2)]).ravel()
hog_feats = hog(arr,
orientations=9,
pixels_per_cell=(8, 8),
cells_per_block=(1, 1),
block_norm='L2-Hys',
feature_vector=True)
ent = entropy(arr)
# # prepare filter bank kernels
# kernels = []
# for theta in range(4):
# theta = theta / 4. * np.pi
# for sigma in (1, 3):
# for frequency in (0.05, 0.25):
# kernel = np.real(gabor_kernel(frequency, theta=theta,
# sigma_x=sigma, sigma_y=sigma))
# kernels.append(kernel)
# gabor_feat = compute_feats(arr, kernels).ravel()
thresh_sauvola = threshold_sauvola(arr, window_size=31, k=0.2)
arr = arr > thresh_sauvola
arr = (255 - arr * 255).astype('uint8')
# arr = adadoc.adath(arr, method=adadoc.ADATH_SAUVOLA | adadoc.ADATH_INVTHRESH,
# xblock=21, yblock=21, k=0.2, dR=64, C=0)
lbp_code = local_binary_pattern(arr, n_points, radius, 'uniform')
# n_bins = int(lbp_code.max() + 1)
n_bins = 16
lbp_feats, _ = np.histogram(lbp_code,
normed=True,
bins=n_bins,
range=(0, n_bins))
data_feat = np.hstack([lbp_feats, ent, glcm_feats, hog_feats])
return data_feat
def threshold_sauvola(arr1d):
"""
doesnt work: TypeError: ndarray() missing required argument 'shape' (pos 1)
:param arr1d:
:return:
"""
import skimage.filters as sf
thresh = sf.threshold_sauvola(arr1d, window_size=15, k=0.2, r=None)
return thresh
def binarization(image):
binarized_image = copy.deepcopy(image)
window_size = 33 #Parameter window_size determines the size of the window that contains the surrounding pixels, window size must be odd and greater than 1
thresh_matrix = threshold_sauvola(
image, window_size=window_size) #Return matrix of thresholds
sauvola_image = image > thresh_matrix #Comparing pixel value with the threshold
binarized_image[sauvola_image == False] = 0
binarized_image[sauvola_image == True] = 255
return binarized_image
def estimate_dominant_fontsize(img, dark_text=None, verbose=0):
ih, iw = img.shape[:2]
kernel = np.array([1, 2, 3, 4, 5, 4, 3, 2, 1], dtype=np.float32)
kernel = kernel / np.sum(kernel)
paper_name, _ = parse_paper_size(ih, iw)
est_lh = compute_fontsize_in_pixels_for_paper(max(ih, iw),
font_size=10,
paper_name=paper_name)
# binarize image
if (img.ndim == 2):
gimg = img
else:
gimg = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
#mimg = threshold_otsu( gimg )
mimg = threshold_sauvola(gimg,
window_size=min(
min(ih, iw) // 20 * 2 + 1,
int(est_lh * 5) // 2 * 2 + 1))
if (dark_text is not None):
if (dark_text):
text = (gimg < mimg).astype('uint8') * 255
else:
text = (gimg > mimg).astype('uint8') * 255
else:
bimg = (gimg < mimg)
if (np.sum(bimg) < bimg.size // 2):
text = bimg.astype('uint8') * 255
verbose_print(
verbose,
"INFO: estimated case = DARK TEXT on bright background")
else:
text = (1 - bimg.astype('uint8')) * 255
verbose_print(
verbose,
"INFO: estimated case = BRIGHT TEXT on dark background")
cvCCA = cv2.connectedComponentsWithStats(text, 8, cv2.CV_32S)
num_regs, labels, reg_stats, centroids = cvCCA
# estimate dominated fontsize
rel_lut = dict()
for label in range(num_regs):
# retrieving the width of the bounding box of the component
width = reg_stats[label, cv2.CC_STAT_WIDTH]
# retrieving the height of the bounding box of the component
height = reg_stats[label, cv2.CC_STAT_HEIGHT]
if (height > est_lh * .75) and (height < est_lh * 10):
if (height not in rel_lut):
rel_lut[height] = 0
rel_lut[height] += width
keys = list(rel_lut)
vals = list(rel_lut.values())
vall = np.sum(vals)
obs = np.zeros(np.max(keys) + 1)
for k, v in zip(keys, vals):
obs[k] = float(v) / vall
n_obs = np.convolve(obs, kernel, mode='same')
dominant_fontsize = np.argmax(n_obs)
return dominant_fontsize
import matplotlib.pyplot as plt
from skimage.data import page
from skimage.filters import (threshold_otsu, threshold_niblack,
threshold_sauvola)
matplotlib.rcParams['font.size'] = 9
image = page()
binary_global = image > threshold_otsu(image)
window_size = 25
thresh_niblack = threshold_niblack(image, window_size=window_size, k=0.8)
thresh_sauvola = threshold_sauvola(image, window_size=window_size)
binary_niblack = image > thresh_niblack
binary_sauvola = image > thresh_sauvola
plt.figure(figsize=(8, 7))
plt.subplot(2, 2, 1)
plt.imshow(image, cmap=plt.cm.gray)
plt.title('Original')
plt.axis('off')
plt.subplot(2, 2, 2)
plt.title('Global Threshold')
plt.imshow(binary_global, cmap=plt.cm.gray)
plt.axis('off')