OpenCV Python Program to blur an image?

OpenCV is one of the best Python packages for image processing. Like signals carry noise attached to them, images too contain different types of noise mainly from the source itself (camera sensor). Python's OpenCV package provides ways for image smoothing, also called blurring. One of the most common techniques is using a Gaussian filter for image blurring, which smooths sharp edges while minimizing excessive blur.

Syntax

cv2.GaussianBlur(src, ksize, sigmaX[, dst[, sigmaY[, borderType]]])

Parameters

• src − Input image

• ksize − Gaussian kernel size as (width, height). Both values must be odd and positive

• sigmaX − Kernel standard deviation along X-axis (horizontal direction)

• sigmaY − Kernel standard deviation along Y-axis (vertical direction). If 0, it equals sigmaX

• dst − Output image (optional)

• borderType − Pixel extrapolation method for image boundaries

Example 1: Basic Gaussian Blur

Here's a simple example that applies Gaussian blur with a small kernel size ?

import cv2
import numpy as np

# Create a sample image with text
img = np.ones((200, 400, 3), dtype=np.uint8) * 255
cv2.putText(img, 'OpenCV Blur Example', (50, 100), cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 0, 0), 2)

# Apply Gaussian blur with small kernel
blurred = cv2.GaussianBlur(img, (5, 5), 0)

# Display both images
cv2.imshow('Original vs Blurred', np.hstack([img, blurred]))
cv2.waitKey(0)
cv2.destroyAllWindows()

Example 2: Different Blur Intensities

Let's compare different kernel sizes to see their effect on blurring ?

import cv2
import numpy as np

# Read an image (replace with your image path)
img = cv2.imread('image.jpg')

# Apply different levels of blur
blur_small = cv2.GaussianBlur(img, (5, 5), 0)
blur_medium = cv2.GaussianBlur(img, (15, 15), 0)
blur_large = cv2.GaussianBlur(img, (31, 31), 0)

# Create a comparison grid
top_row = np.hstack([img, blur_small])
bottom_row = np.hstack([blur_medium, blur_large])
result = np.vstack([top_row, bottom_row])

cv2.imshow('Blur Comparison', result)
cv2.waitKey(0)
cv2.destroyAllWindows()

Example 3: Using Different Sigma Values

You can control blur intensity using sigma values instead of relying on kernel size ?

import cv2
import numpy as np

# Create a sample image
img = np.zeros((300, 400, 3), dtype=np.uint8)
cv2.rectangle(img, (50, 50), (350, 250), (255, 255, 255), -1)
cv2.rectangle(img, (100, 100), (300, 200), (0, 0, 255), -1)

# Apply blur with different sigma values
blur_sigma_1 = cv2.GaussianBlur(img, (0, 0), sigmaX=1)
blur_sigma_3 = cv2.GaussianBlur(img, (0, 0), sigmaX=3)
blur_sigma_5 = cv2.GaussianBlur(img, (0, 0), sigmaX=5)

# Display results
result = np.hstack([img, blur_sigma_1, blur_sigma_3, blur_sigma_5])
cv2.imshow('Different Sigma Values', result)
cv2.waitKey(0)
cv2.destroyAllWindows()

Key Points

• Kernel size must be odd numbers (3, 5, 7, 15, etc.)

• Larger kernel sizes create more blur effect

• When ksize is (0,0), kernel size is calculated from sigma values

• Higher sigma values produce more pronounced blurring

• Gaussian blur preserves edges better than simple averaging

Common Use Cases

• Noise reduction in images

• Pre-processing for edge detection

• Creating background blur effects

• Smoothing images for better feature extraction

Conclusion

Gaussian blur is an essential image processing technique in OpenCV that effectively reduces noise while preserving important image features. By adjusting kernel size and sigma values, you can control the blur intensity to suit your specific application needs.