猜猜你的意思，这里有一个方法：

• 加载图像为灰度
• 纯黑和白色阈值
• 获取所有黑色像素的坐标
• 随机抽取0.5%的坐标
• 创建空输出图像
• 在二次采样位置的输出图像的alpha通道上绘制白色圆圈，这将使黑色背景在这些位置变得可见

#!/usr/bin/env python3

import numpy as np
import cv2 as cv

# Load image as greyscale
im = cv.imread('Gkzaa.png', cv.IMREAD_GRAYSCALE)

# Otsu threshold to pure black and pure white, i.e. 0 or 255
_, thresh = cv.threshold(im, 0, 255, cv.THRESH_BINARY+cv.THRESH_OTSU)

# Get X and Y coordinates of all black pixels
Y, X = np.where(thresh==0)
nBlack = len(Y)

# Get indices of 0.5% of those black pixels
indices = np.random.choice(nBlack, size=int(nBlack*0.005), replace=False)

# Form an empty (black) output image and an alpha channel for it
BGR = np.zeros((*im.shape,3), np.uint8)
A   = np.zeros_like(im)

# Draw circles of opacity into the alpha channel at selected indices
for i in indices:
    cv.circle(A, center=(X[i], Y[i]), radius=4, color=255, thickness=cv.FILLED)

# Stack the alpha channel onto the BGR channel to make BGRA and save
res = np.dstack((BGR,A))
cv.imwrite('result.png', res)