如何使用Python处理图像

Question

我有一张图片。这是一封书法字母。

我需要让它充满点，使所有的点相互跟随并对应于字母线的粗细，也就是说，如果字母线变粗，那么点就变大。这是一个例子：

（原来是这封信）

我尝试使用 OpenCV 和 Pillow 处理照片的骨架，但没有成功，如果您帮助我，我将非常高兴，提前谢谢您！

Answer 1

玩了一下，得出了这种相当幼稚的方法，它看起来迭代地放置点，并且似乎产生的结果在某种程度上符合我相信你正在寻找的方向：

import matplotlib.pyplot as plt
from PIL import Image
import numpy as np

# read image into numpy array
img = np.asarray(Image.open('../zZfpE.jpg'))
# binarize and coarse-graining
img = (img[::2, ::2, 0] < 128).astype(int)

def make_dots(
    img, # 2d np array of 1s and 0s
    spd=25, # speed up factor
    marg=0.5, # minimum spacing between dots relative to dot size,
    max_ds=27, # maximum dot diameter in pixels
    min_ds=6, # minimum dot diameter in pixels    
):

    # start with blank image
    img2 = np.zeros(img.shape)
    h, w = img.shape

    # keep track of areas occupied by dots
    occ = set()
    
    ofs = int(spd ** 0.5)

    # iterate over dot sizes, decreasing
    for t in np.arange(max_ds, min_ds, -5):
        # iterate over pixels
        for i in range(0, h, ofs):
            for j in range(0, w, ofs):
                d = 0
                # increase dot diameter while 1. within bounds, b. corners of bounding box are black and 3. not already occupied
                while (
                    i + d < h and j + d < w
                ) and all(
                    [img[i + d, j + d], img[i, j + d], img[i + d, j]]
                ) and not any(
                    [(i + d1, j + d2) in occ for d1, d2 in [(0, d), (d, 0), (d, d)]]
                ):
                    d += 1
                # consider dot only if exceeding threshold
                if d > t:
                    m = int(marg * d)
                    ci = i + d // 2
                    cj = j + d // 2
                    d2 = (d//2) ** 2
                    # iterate over pixels in bounding box of dot + margin
                    for a in range(i-m, i+d+m):
                        for b in range(j-m, j+d+m):
                            # mark pixel as occupied by dot
                            occ.add((a, b))
                            # if within radius, draw to image
                            if (ci - a) ** 2 + (cj - b) ** 2 < d2:
                                img2[a, b] = 1
    return img2


plt.imshow(np.hstack((img, make_dots(img))), cmap='gray_r')
plt.axis('off')
plt.tight_layout()

虽然它有点脆弱/老套，但它（根据手头的数据定制参数，或其修改版本）可能仍然有用，如果没有，至少希望能提供一些灵感，因此在这里分享。