Search code examples
pythonimage-processingscikit-image

How to extract a smooth skeleton from an image

I have some font character images of fixed size as shown under the Input image sample. I want to extract the character skeleton (single-pixel wide). I have tried various ways as shown below but the outputs are all different and not smooth. I thought that the one-pixel-wide skeleton will be smooth (pixels not breaking and no noise pixels). Is there a better way to do this? If no which one is the best among these three?

Input image sample

enter image description here

1) Example

from skimage import img_as_bool, io, color, morphology
import matplotlib.pyplot as plt

image = img_as_bool(color.rgb2gray(io.imread('image.jpeg')))
out = morphology.medial_axis(image)

f, (ax0, ax1) = plt.subplots(1, 2)
ax0.imshow(image, cmap='gray', interpolation='nearest')
ax1.imshow(out, cmap='gray', interpolation='nearest')
plt.show()

Output1

enter image description here

2) Example

from PIL import Image, ImageDraw, ImageFont
import mahotas as mh
import numpy as np

image = Image.new("RGBA", (600,150), (255,255,255))
draw = ImageDraw.Draw(image)
fontsize = 150
font = ImageFont.truetype("font.TTF", fontsize)
txt = '가'
draw.text((30, 5), txt, (0,0,0), font=font)
img = image.resize((188,45), Image.ANTIALIAS)
print(type(img))
plt.imshow(img)

img = np.array(img)
im = img[:,0:50,0]
im = im < 128
skel = mh.thin(im)
noholes = mh.morph.close_holes(skel)
plt.subplot(311)
plt.imshow(im)
plt.subplot(312)
plt.imshow(skel)

Output2

enter image description here

3) Example 

from skimage.morphology import skeletonize
from skimage import draw
from skimage.io import imread, imshow
from skimage.color import rgb2gray
import os

# load image from file
img_fname='D:\Ammar Data\Debbie_laptop_data\Ammar\sslab-deeplearning\GAN models\sslab_GAN\skeleton\hangul_1.jpeg' 
image=imread(img_fname)

# Change RGB color to gray 
image=rgb2gray(image)

# Change gray image to binary
image=np.where(image>np.mean(image),1.0,0.0)

# perform skeletonization
skeleton = skeletonize(image)

plt.imshow(skeleton)

output3

enter image description here

Solution

  • Your code is fine but you may need to change the way you are converting the image to binary. Also, to avoid the noisy looking output, you can apply binary_closing to your skeleton image. Take a look at the below code -

    import matplotlib.pyplot as plt
from skimage import img_as_bool
from skimage.io import imread
from skimage.color import rgb2gray
from skimage.morphology import skeletonize, binary_closing


im = img_as_bool(rgb2gray(imread('0jQjL.jpg')))
out = binary_closing(skeletonize(im))

f, (ax0, ax1) = plt.subplots(1, 2)
ax0.imshow(im, cmap='gray', interpolation='nearest')
ax1.imshow(out, cmap='gray', interpolation='nearest')
plt.show()

    Your two sample images gave me the below output -

    enter image description here

    enter image description here

    EDIT: To avoid the precision loss when converting the image to bool, you can also binarize the image using one of the available thresholding algorithms. I prefer otsu's.

    import matplotlib.pyplot as plt
from skimage.io import imread
from skimage.filters import threshold_otsu
from skimage.color import rgb2gray
from skimage.morphology import skeletonize, binary_closing

def get_binary(img):    
    thresh = threshold_otsu(img)
    binary = img > thresh
    return binary

im = get_binary(rgb2gray(imread('Snip20190410_9.png')))
out = binary_closing(skeletonize(im))

f, (ax0, ax1) = plt.subplots(1, 2)
ax0.imshow(im, cmap='gray', interpolation='nearest')
ax1.imshow(out, cmap='gray', interpolation='nearest')
plt.show()