使用OpenCV Python查找字母并将其上色为红色

Question

我有一个很大的图像，里面有一些字母，并且切出一个字母（“ A”）。我需要在较大的图像中找到每个A，然后将其上色为红色。

大图：

字母A：

为解决此问题，我使用了以下代码-

import cv2, numpy as np

# read the image and convert into binary
a = cv2.imread('search.png', 0) 
ret,binary_image = cv2.threshold(a,230,255,cv2.THRESH_BINARY_INV)

# create the Structuring element
letter_a = cv2.imread('A.png', 0)
ret,se = cv2.threshold(letter_a,230,255,cv2.THRESH_BINARY_INV)

#erosion and dilation for finding A
erosion = cv2.erode(binary_image , se) 
new_se = cv2.flip(se,0)
dilation = cv2.dilate(erosion, new_se) 
cv2.imwrite('dilation.jpg', dilation )

在这一点上，我得到以下图像

如您所见，我清楚地标识了所有A。但是，我需要将A涂成红色，最重要的是，在第一个大图像上写上黑色字母和白色背景。有什么办法吗？也许在第一个图像上使用numpy数组写？

Answer 1

您可以按照以下方法解决此问题。
首先，要将主图像中的字母涂成红色，最好将其加载成彩色。创建灰度副本以执行阈值。
 然后创建具有主图像尺寸的黑色图像，并将该图像的颜色设置为红色。带A的图像用作蒙版，以获取红色A的图像。然后将这些红色的A添加到主图像中。*

结果：

代码：

import cv2, numpy as np

# load the image in color
a = cv2.imread('search.png') 
# create grayscale
a_gray = cv2.cvtColor(a,cv2.COLOR_BGR2GRAY)
ret,binary_image = cv2.threshold(a_gray,230,255,cv2.THRESH_BINARY_INV)

# create the Structuring element
letter_a = cv2.imread('A.png', 0)
ret,se = cv2.threshold(letter_a,230,255,cv2.THRESH_BINARY_INV)

#erosion and dilation for finding A
erosion = cv2.erode(binary_image , se) 
new_se = cv2.flip(se,0)
dilation = cv2.dilate(erosion, new_se) 

# create a red background image
red = np.zeros((a.shape[:3]),dtype=a.dtype)
red[:] = (0,0,255)
# apply the mask with A's to get red A's
red_a = cv2.bitwise_and(red,red,mask=dilation)

# Add the A's to the main image
result = cv2.add(a,red_a)

cv2.imshow('Result', result )
cv2.waitKey(0)
cv2.destroyAllWindows()

*如果字母不是黑色，则需要额外的步骤，请阅读此tutorial。但是对于您的图像，这不是必需的。

Answer 2

我使用以下代码解决了问题-

import cv2, numpy as np
# read the image and convert into binary
color_image = cv2.imread(r'search.png', 1) 
gray_image = cv2.imread(r'search.png', 0) 
ret,binary_image = cv2.threshold(gray_image,230,255,cv2.THRESH_BINARY_INV)

# create the Structuring element
letter_a = cv2.imread('A.png', 0)
ret,se = cv2.threshold(letter_a,230,255,cv2.THRESH_BINARY_INV)

#erosion and dilation for finding A
erosion = cv2.erode(binary_image, se) 
new_se = cv2.flip(se,0)
dilation = cv2.dilate(erosion, new_se) 

for i in zip(*np.where(dilation == 255)):
    color_image[i[0], i[1], 0] = 0
    color_image[i[0], i[1], 1] = 0
    color_image[i[0], i[1], 2] = 255

# show and save image
cv2.imwrite('all_a.jpg', color_image)
cv2.imshow('All A',color_image)
cv2.waitKey(0)
cv2.destroyAllWindows()