Opencv learning note 5 - gradient calculation / edge detection

To get the outline , An image can be expanded first and then corroded , Then the result of expansion minus the result of corrosion .

1. Sobel operator

dst = cv2.Sobel(src, ddepth, dx, dy, ksize)

• ddepth: The depth of the image
• dx nucleus dy Horizontal and vertical directions respectively
• ksize yes Sobel Operator size

image1 = cv2.imread('./ro.png')
plt.imshow(image1);

image1 = cv2.imread('./ro.png')
plt.imshow(image1);

image2 = cv2.Sobel(image1, cv2.CV_64F, 1, 0, ksize=3)
plt.imshow(image2);

The convolution calculation from white to black is positive , The convolution calculation from black to white is negative , All negative numbers will be truncated to 0, So take the absolute value

Calculation x Direction gradient

image2 = cv2.Sobel(image1, cv2.CV_64F, 1, 0, ksize=3)
image2 = cv2.convertScaleAbs(image2)
plt.imshow(image2);

Calculation x Direction gradient

image3 = cv2.Sobel(image1, cv2.CV_64F, 0, 1, ksize=3)
image3 = cv2.convertScaleAbs(image3)
plt.imshow(image3);

Get the outline of the image

image4 = cv2.addWeighted(image2, 0.5, image3, 0.5, 0)
plt.imshow(image4);

Try another picture

import cv2
import matplotlib.pyplot as plt
image = cv2.imread('image.png')
image2 = cv2.Sobel(image, cv2.CV_64F, 1, 0, ksize=3)
image2 = cv2.convertScaleAbs(image2)
image3 = cv2.Sobel(image, cv2.CV_64F, 0, 1, ksize=3)
image3 = cv2.convertScaleAbs(image3)
image4 = cv2.addWeighted(image2, 0.5, image3, 0.5, 0)
image4 = cv2.cvtColor(image4, cv2.COLOR_BGR2RGB)
plt.imshow(image4);

2. Scharr operator

The usage is the same , But the calculated value will be relatively large , Will be more sensitive

3. laplacian operator