如何得到图的每个像素的x,y？

Question

我有以下图片：

我需要完成 2 个步骤。

1 步。 获取图形每个像素的值 (x, y)。结果：图形的数量等于图形的每个像素包含对象 {'x': x, 'y': y} 的数组的数量。 https://docs.scipy.org/doc/scipy/reference/generated/scipy.signal.find_peaks.html

第 2 步。 每个图的每个像素数组都需要在 scipy.find_peaks 中运行并获取峰值（每个图的最大点数）。

from PIL import Image
import cv2

img_gray = cv2.imread('photo.jpg', cv2.IMREAD_GRAYSCALE)
img_binary = cv2.threshold(img_gray, 150, 255, cv2.THRESH_BINARY)[1]
cv2.imwrite('save.jpg', img_binary)

image = cv2.imread('save.jpg')
cv2.imwrite('save.jpg', image)


im = Image.open("save.jpg")
im = im.convert('RGB')

col_black = (0, 0, 0)
col_white = (255, 255, 255)

for y in range(im.size[1]):
    for x in range(im.size[0]):
        if sum(im.getpixel((x, y))) < 128:
            im.putpixel((x, y), (0, 0, 0))
        else:
            im.putpixel((x, y), (255, 255, 255))

point_start_array = []
for x in range(im.size[0]):
    for y in range(im.size[1]):
        if im.getpixel((x, y)) == col_black and im.getpixel((x, y + 1)) == col_white:
            if len(point_start_array) > 4:
                break
            else:
                point_start_array.append({'x': x, 'y': y})

print(point_start_array)

输出： [{'x': 19, 'y': 161}, {'x': 19, 'y': 238}, {'x': 19, 'y ': 260}, {'x': 19, 'y': 269}, {'x': 19, 'y': 282}]

是否有可能每个图抛出 1 个数组 x，第二个 y 抛出 scipy.find_peaks 并获得峰值？

Answer 1

您可以使用 opencv connectedComponents 来获取图形的点和一些 numpy 技巧来获取 y 数组。

import cv2
import numpy as np
import scipy
from scipy.signal import find_peaks

im = cv2.imread('temp.png')
im = cv2.cvtColor(im,cv2.COLOR_BGRA2GRAY)
binary = cv2.threshold(im, 0, 255, cv2.THRESH_BINARY | cv2.THRESH_OTSU)[1]
binary = ~binary
# Get connected components
ret, labels = cv2.connectedComponents(binary)
graphs = []
for label in range(1,ret):
    mask = np.array(labels, dtype=np.uint8)
    mask[labels == label] = 255
    mask = mask.reshape((mask.shape[0], mask.shape[1], 1))
    X,Y = np.where(np.all(mask == 255,axis=2))
    graphs.append(Y)

for ys in graphs:
  peaks, _ = find_peaks(ys, height=0)

但请注意，图像中的一个图形不是完全连接的，即：缺少一个像素来连接图形。所以不是 5 个，而是 6 个图表。我建议使用腐蚀等图像处理技术来连接图形，但会牺牲一些准确性

import cv2
import numpy as np
import scipy
from scipy.signal import find_peaks

im = cv2.imread('temp.png')
im = cv2.cvtColor(im,cv2.COLOR_BGRA2GRAY)
binary = cv2.threshold(im, 0, 255, cv2.THRESH_BINARY | cv2.THRESH_OTSU)[1]
# Erode the images to connect seperated parts
kernel = np.ones((3, 3), np.uint8)  
binary = cv2.erode(binary, kernel, cv2.BORDER_REFLECT) 
# Get connected components
binary = ~binary
ret, labels = cv2.connectedComponents(binary)
graphs = []
for label in range(1,ret):
    mask = np.array(labels, dtype=np.uint8)
    mask[labels == label] = 255
    cv2.imwrite(f'{label}.jpg',mask)
    mask = mask.reshape((mask.shape[0], mask.shape[1], 1))
    X,Y = np.where(np.all(mask == 255,axis=2))
    graphs.append(Y)
for ys in graphs:
  peaks, _ = find_peaks(ys, height=0)