def line_seg(img):
img_not = abs(255 - img)
#horizontal histogram
histo = img_not.sum(axis=1)
#smoothening histogram for better detection of lines
y_smooth = savgol_filter(histo, 27, 3)
#comparing histogram after smoothening
# plt.plot(histo)
# plt.plot(y_smooth)
minimas = argrelextrema(y_smooth, np.less)
arr = minimas[0]
# print(arr)
# plotting minimas in histogram i.e. line breaks
# for i in arr:
# plt.plot(i, y_smooth[i], 'ro')
# plt.show()
#creating a list of lines
lines = []
prev = 0
for i in arr:
if (prev != -1):
curr = img_not[prev:i + 1, :]
if (np.sum(curr) >= 6 * 255):
curr = img[prev:i + 1, :]
curr = pre.cut_image(curr)
lines.append(curr)
prev = i
i = len(histo) - 1
curr = img_not[prev:i + 1, :]
if (np.sum(curr) >= 6 * 255):
curr = img[prev:i + 1, :]
curr = pre.cut_image(curr)
lines.append(curr)
return lines
def pre_init(img):
img = pre.cut_image(img)
aspect = img.shape[0] / img.shape[1]
img = cv2.resize(img, (50, 50), interpolation=cv2.INTER_CUBIC)
img = pre.otsu_thresh(img)
new_img = pre.thin_image(img)
# changes for end-points counting
# img = util.cnvt_bool_to_uint8(img)
# kernel = np.ones((3,3),np.uint8)
# dilation = cv2.dilate(img, kernel, iterations=1)
# dilation[dilation==255]=1
return aspect, img, new_img
def get_granule_count(processed_img: np.array) -> (int, np.array):
"""
Returns approximate count of circles in image,
and distribution of circles' radiuses
"""
proc = cut_image(processed_img, 30)
resized = cv2.resize(proc, (AVG_W, AVG_H))
avg_r, circles = draw_hough(resized)
found = count_circles(proc, avg_r)
radiuses = np.array([circle[2] for circle in circles])
distros = np.array(
[np.where(radiuses == i)[0].shape[0] for i in range(1, 30)])
return found, distros / np.sum(distros)
def get_stats(im: np.array) -> (np.array, np.array):
"""
Returns stats and areas of connected components of the image
Args:
im: np.array
Returns:
stats: np.array - stats of every component
label_area:np.array - area of every component
"""
img = cv2.resize(im, (hough.AVG_W, hough.AVG_H))
img = preprocess.cut_image(img, 30)
mask = cv2.threshold(
img[:, :, 0], 255, 255, cv2.THRESH_BINARY_INV + cv2.THRESH_OTSU
)[1]
stats = cv2.connectedComponentsWithStats(mask, 4)[2]
label_area = stats[1:, cv2.CC_STAT_AREA]
return stats, label_area
def cut_images_save(img,
if_show_pre=False,
if_show=False,
img_name='',
save_path=''):
global image_index
print('**********************开始处理图片*************************')
print('img_name: ', img_name)
wrap = detect_image_counts(img, if_show_pre, if_show, img_name)
print('wrap: ', wrap)
if wrap == 2:
print('-------------此图片有两张发票----------------')
imgs_list = cut_image(img)
for img in imgs_list:
cv2.imwrite(save_path + str(image_index) + '.png', img)
image_index += 1
else:
cv2.imwrite(save_path + str(image_index) + '.jpg', img)
image_index += 1
print(
'***************************************处理图片完成*************************************'
)