How to smooth a skeletonized image and search for a point on the smoothed line?-CodePudding

I have a skeletonized image and I want to obtain a smoothened one from the image attached here below. The main goal after smoothening is to locate any point nearby or exactly at the center of the horizontal plane, I just want to pick any point say (x, y) = (530, 355) and determine if it lies on the horizontal line whether exactly at the center or at any part of the skeleton. Any procedures to go through this?

My trial

I tried to find the contour and then determine the largest contour from the skeleton but I cannot determine any point that lies on the horizontal plane. Any idea to go through this

CodePudding user response：

If I understand you correctly you are interested in the diagonal of the skeleton. Here is one idea that may need some work before it does exactly what you want.

I want to find a line where the picture is white by optimising the length and brightness of some starting line. For that it would be useful if the skeleton would be a bit thicker and it would turn into black slowly away from the center. So I do the following preprocessing

import numpy as np
import matplotlib.pyplot as plt
import cv2
from scipy.ndimage import map_coordinates

img = cv2.imread('test.png', cv2.IMREAD_GRAYSCALE)
img = ((img>100)*255).astype('uint8')
n = 20

img = cv2.dilate(img, np.ones((n 5,n 5)))
img = cv2.blur(img, (n 1,n 1))

Now I do the optimisation mentioned above

def get_points_on_line(x):
    p = x[:2].reshape(-1,1)
    q = x[2:].reshape(-1,1)
    t = np.linspace(0,1,10**3)
    coords = t*p (1-t)*q
    length = np.linalg.norm(p-q)
    values = (map_coordinates(img, coords, prefilter=False)-255*9/10)
    value = values.mean()
    return -value*length

res = minimize(get_points_on_line, [250,200,500,700], method = 'Nelder-Mead')

and plot the result

x = res.x
p = x[:2].reshape(-1,1)
q = x[2:].reshape(-1,1)
t = np.linspace(0,1,10**3)
coords = t*p (1-t)*q

plt.imshow(img, cmap='gray')
plt.plot(coords[1], coords[0])