Image filtering (gaussian_filter) in SciPy

Introduction

We use image filtering to smooth images and reduce noise. The gaussian_filter helps blur an image gently by averaging nearby pixels with a bell-shaped curve.

To remove small specks or noise from a photo taken in low light.

To prepare an image before detecting edges or shapes.

To create a soft blur effect for artistic purposes.

To smooth data in scientific images like medical scans.

To reduce detail before compressing an image.

Syntax

SciPy

scipy.ndimage.gaussian_filter(input, sigma, order=0, output=None, mode='reflect', cval=0.0, truncate=4.0)

input: The image or array to filter.

sigma: Controls how much to blur. Bigger sigma means more blur.

Examples

Apply a mild blur with sigma 1 to the image.

SciPy

from scipy.ndimage import gaussian_filter

blurred = gaussian_filter(image, sigma=1)

Apply a stronger blur with sigma 3 for more smoothing.

SciPy

blurred = gaussian_filter(image, sigma=3)

Apply different blur amounts for each axis in a 2D image.

SciPy

blurred = gaussian_filter(image, sigma=[1, 2])

Sample Program

This code creates a small noisy image with a bright square in the center. Then it smooths the image using gaussian_filter with sigma=1. Finally, it prints both the original and blurred images as arrays rounded to 2 decimals.

SciPy

import numpy as np
from scipy.ndimage import gaussian_filter
import matplotlib.pyplot as plt

# Create a simple 2D image with noise
np.random.seed(0)
image = np.zeros((10, 10))
image[4:6, 4:6] = 10  # bright square in the middle
image += np.random.normal(0, 1, image.shape)  # add noise

# Apply gaussian filter with sigma=1
blurred_image = gaussian_filter(image, sigma=1)

# Print original and blurred images
print('Original image:\n', np.round(image, 2))
print('\nBlurred image:\n', np.round(blurred_image, 2))

Important Notes

The sigma value controls the blur strength. Try small values first.

Gaussian filter works well to reduce noise but keeps edges smoother than simple averaging.

Use mode parameter to control how edges are handled (default is 'reflect').

Summary

Gaussian filter smooths images by averaging pixels with a bell curve.

It helps reduce noise and prepare images for further analysis.

Adjust sigma to control how much blur you want.