CODEWITHSUNDEEP
opencvimage-processingmathematical-morphology
SaulGoodMan
SaulGoodMan

Reputation: 173

Rounded corners dilation (Image Processing)

I want to perform an Dilation operation while having rounded corners. Something like this : Example

What I tried :

import numpy as np
import cv2

img = cv2.imread(test.jpg)
kernel = np.array([[0,1,0],
                   [1,1,1],
                   [0,1,0]], dtype=np.uint8)
img_d = cv2.dilate(img, kernel, iterations=45)

Image used : test.jpg
I tried multiple kernels (with different sizes 3x3, 5x5...) but I didn't succeed to get rounded corners.

enter image description here

My Question is : Can we get rounded corners just by changing the kernel or should we add a further processing step to achieve this ?


NOTE : My goal is not to create a rounded square... I used this example just to explain the idea of getting rounded corners with a dilation operation.

Upvotes: 4

Views: 1430

Answers (2)

Jeru Luke
Jeru Luke

Reputation: 21203

OpenCV also allows you choose a kernel of your shape and size with cv2.getStructuringElement. From this page you can choose either a rectangle, ellipse or cross shaped kernels.

Since you needed rounded corners I chose the ellipse kernel cv2.MORPH_ELLIPSE of size 25 x 25:

img = cv2.imread('test.jpg', 0)
th_img = cv2.threshold(gray,127,255,cv2.THRESH_BINARY)[1]  

kernel =  cv2.getStructuringElement(cv2.MORPH_ELLIPSE,(25,25))
img_dil = cv2.dilate(th_img, kernel, iterations=5)

Not exactly the result you were hoping for but here is how it looks:

enter image description here

Upvotes: 2

SaulGoodMan
SaulGoodMan

Reputation: 173

I figured out a way thanks to "Christoph Rackwitz" comment.

The idea is pretty simple. We need to use a bigger kernel with a circle shape and reduce the number of Dilation iterations. 

import numpy as np
import cv2

kernel = np.zeros((100,100), np.uint8)
cv2.circle(kernel, (50,50), 50, 255, -1)
plt.imshow(kernel, cmap="gray")

circle

And then use this kernel with just one iteration :

img = cv2.imread(test.jpg)
img_d = cv2.dilate(img, kernel, iterations=1)
plt.imshow(kernel, cmap="gray")

output

Upvotes: 5

Related Questions