Sunday 17 December 2023

OpenCV: Sobel edge detection

Sobel operation is used for edge detection and gradient computation in Image processing.

 

Sobel operation has two 3X3 convolution kernels.

a.   Kernel1 : For detecting changes in intensity in the horizontal (x) direction

b.   Kernel2: For detecting the changes in vertical direction.

 

Horizontal kernel

-1  0  1

-2  0  2

-1  0  1

 

Vertical Kernel

-1 -2 -1

 0  0  0

1     2  1

 

How to find edges using Sobel operator in OpenCV?

Using cv2.Sobel method, we can apply Sobel operator on an Image.

 

Signature

cv2.Sobel(src, ddepth, dx, dy, ksize[, scale[, delta[, borderType]]])

 

Following table summarizes the parameters of Sobel method.

Parameter

Description

src

Specifies the source image on which you want to apply sobel operator.

ddepth

Specifies desired depth of the image. By default it is set to -1 to indicate output image has same depth as source image.

dx

Order of derivatives in x-axis

dy

Order of derivatives in y-axis

ksize

Specifies the size of Sobel kernel. In general it is set to odd number like 3, 5, 7 etc.,

scale

It is an optional parameter used to scale the result

delta

It is an optional parameter, added to the result.

borderType

It is optional parameter, specifies how to handle border pixels.

 

sobel_edge_detection.py

import cv2 as cv
import numpy as np

image = cv.imread('gateway_of_india.png')
gray_scale_image = cv.cvtColor(image, cv.COLOR_BGR2GRAY)

sobel_x = cv.Sobel(image, -1, 1, 0)
sobel_y = cv.Sobel(image, -1, 0, 1)
sobel = cv.bitwise_or(sobel_x, sobel_y)

cv.imshow('Original image', image)
cv.imshow('gray_scale_image', gray_scale_image)
cv.imshow('sobel_x', sobel_x)
cv.imshow('sobel_y', sobel_y)
cv.imshow('sobel', sobel)

cv.waitKey(0)

# Close the OpenCV windows
cv.destroyAllWindows()

 

Output

Original image

 


 

Gray scale image

 


sobel_x




sobel_y

 


 

sobel



Previous                                                    Next                                                    Home

No comments:

Post a Comment