从图像生成无噪声的二值图像和骨架

Question

我有一个重叠的细丝图像。我感兴趣的是生成无噪音的二值图像，然后使用它生成骨架。我尝试过不同的方法来获得骨架，但没有成功。下面找到用python编写的相同代码，并附带通过它生成的骨架映像。如果有人帮忙解决问题，那就太好了。

原始图像与骨架图像：

​

​

import cv2
import numpy as np
from skimage import morphology, graph
from skan import Skeleton
from skimage.morphology import skeletonize
import matplotlib.pyplot as plt

img00 = cv2.imread(r'img_test.jpg')
img01 = cv2.cvtColor(img00, cv2.COLOR_BGR2GRAY)
cv2.imshow('1',img01)

img02 = cv2.adaptiveThreshold(img01,255,cv2.ADAPTIVE_THRESH_GAUSSIAN_C, cv2.THRESH_BINARY,5,5)
cv2.imshow('2',img02)


i_size = min(np.size(img02,1),600) # image size for imshow
kernel = np.ones((2, 2), np.uint8)  # Creating kernel

# Using cv2.erode() method 
img_erosion = cv2.erode(img02, kernel, borderType = cv2.BORDER_REFLECT, iterations=1, borderValue = 1)

filterSize =(5,5)
kernel = cv2.getStructuringElement(cv2.MORPH_RECT, filterSize)
tophat_img = cv2.morphologyEx(img_erosion, cv2.MORPH_BLACKHAT, kernel)
img03 = cv2.bitwise_not(tophat_img)
cv2.imshow('3',img03)

kernel1 = cv2.getStructuringElement(cv2.MORPH_ELLIPSE,(2,2))
img04 = cv2.morphologyEx(img03, cv2.MORPH_CLOSE, kernel1)
img04 = cv2.morphologyEx(img04, cv2.MORPH_OPEN, kernel1)
cv2.imshow('4',img04)

thresh = (img04/255).astype(np.uint8)

# skeleton based on default method
skeleton1 = skeletonize(thresh)
skeleton2 = (skeleton1*255).astype(np.uint8)
cv2.imshow('5',skeleton2)

# Avg diameter calculation
diameter = np.sum(thresh)/np.sum(skeleton1)
print('diameter',diameter)

cv2.waitKey(0) & 0xFF == ord('q')
cv2.destroyAllWindows()

Answer 1

直接二值化和形态学闭合可以给出有趣的结果，尽管它们对参数很敏感。

​

​