Sobel and Laplace edge detection in SciPy

Introduction

Edge detection helps find the outlines of objects in images. Sobel and Laplace methods highlight these edges by showing where colors or brightness change sharply.

To find the borders of objects in a photo for counting or measuring.

To prepare images for further analysis like recognizing shapes or letters.

To detect changes in textures or patterns in medical images.

To improve the clarity of features in satellite images.

To help robots or self-driving cars understand their surroundings.

Syntax

SciPy

import numpy as np
from scipy import ndimage

# Sobel edge detection
sobel_horizontal = ndimage.sobel(image, axis=0)
sobel_vertical = ndimage.sobel(image, axis=1)
sobel_edges = np.hypot(sobel_horizontal, sobel_vertical)

# Laplace edge detection
laplace_edges = ndimage.laplace(image)

image should be a 2D array representing a grayscale image.

Sobel detects edges by looking at horizontal and vertical changes separately, then combines them.

Examples

Detects edges by looking at changes along rows (horizontal edges).

SciPy

sobel_horizontal = ndimage.sobel(image, axis=0)
print(sobel_horizontal.shape)

Detects edges by looking at changes along columns (vertical edges).

SciPy

sobel_vertical = ndimage.sobel(image, axis=1)
print(sobel_vertical.shape)

Detects edges by looking at the second derivative, highlighting areas where brightness changes quickly.

SciPy

laplace_edges = ndimage.laplace(image)
print(laplace_edges.shape)

Sample Program

This code creates a simple image with a white square on black background. It then finds edges using Sobel and Laplace methods. The printed arrays show the edge strength values. The plots help you see where edges are detected.

SciPy

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

# Create a simple 5x5 image with a white square in the middle
image = np.zeros((5, 5))
image[1:4, 1:4] = 1

# Apply Sobel edge detection
sobel_horizontal = ndimage.sobel(image, axis=0)
sobel_vertical = ndimage.sobel(image, axis=1)
sobel_edges = np.hypot(sobel_horizontal, sobel_vertical)

# Apply Laplace edge detection
laplace_edges = ndimage.laplace(image)

# Print arrays to see edge values
print('Original Image:\n', image)
print('\nSobel Edges:\n', sobel_edges)
print('\nLaplace Edges:\n', laplace_edges)

# Plot images for visual understanding
plt.figure(figsize=(10,3))
plt.subplot(1,4,1)
plt.title('Original')
plt.imshow(image, cmap='gray')
plt.axis('off')

plt.subplot(1,4,2)
plt.title('Sobel Horizontal')
plt.imshow(sobel_horizontal, cmap='gray')
plt.axis('off')

plt.subplot(1,4,3)
plt.title('Sobel Vertical')
plt.imshow(sobel_vertical, cmap='gray')
plt.axis('off')

plt.subplot(1,4,4)
plt.title('Laplace')
plt.imshow(laplace_edges, cmap='gray')
plt.axis('off')

plt.tight_layout()
plt.show()

Important Notes

Sobel edges combine horizontal and vertical changes to find strong edges.

Laplace uses second derivatives, so it can detect edges differently, sometimes highlighting corners.

Input images should be grayscale (2D arrays) for these filters to work correctly.

Summary

Sobel and Laplace are simple ways to find edges in images.

Sobel looks at horizontal and vertical changes separately and then combines them.

Laplace looks at how brightness changes twice to find edges.

Practice

1. What is the main purpose of using the Sobel filter in image processing?

easy

A. To detect edges by highlighting horizontal and vertical changes

B. To blur the image and reduce noise

C. To increase the brightness of the image

D. To convert the image to grayscale

Solution

1. Step 1: Understand Sobel filter function

   The Sobel filter detects edges by calculating gradients in horizontal and vertical directions separately.

2. Step 2: Identify the purpose of edge detection

   Edges are found by highlighting where brightness changes sharply, which Sobel does by combining horizontal and vertical gradients.

3. Final Answer:

   To detect edges by highlighting horizontal and vertical changes -> Option A

4. Quick Check:

   Sobel = edge detection [OK]

Hint: Sobel finds edges by checking horizontal and vertical changes [OK]

Common Mistakes:

• Confusing Sobel with blurring filters
• Thinking Sobel changes brightness directly
• Mixing Sobel with color conversion

2. Which of the following is the correct way to apply the Sobel filter on a 2D image array named img using SciPy?

easy

A. scipy.ndimage.sobel(img, axis=2)

B. scipy.ndimage.laplace(img, axis=1)

C. scipy.ndimage.sobel(img, axis=0)

D. scipy.ndimage.gaussian_filter(img, sigma=1)

Solution

1. Step 1: Recall Sobel filter usage in SciPy

   The Sobel filter is applied using scipy.ndimage.sobel with the image and axis (0 for vertical, 1 for horizontal).

2. Step 2: Check axis validity

   For a 2D image, valid axes are 0 or 1. Axis=2 is invalid and will cause an error.

3. Final Answer:

   scipy.ndimage.sobel(img, axis=0) -> Option C

4. Quick Check:

   Sobel syntax = scipy.ndimage.sobel(img, axis) [OK]

Hint: Use axis 0 or 1 for 2D images with sobel [OK]

Common Mistakes:

• Using axis=2 on 2D images causes errors
• Confusing laplace and sobel functions
• Using gaussian_filter instead of sobel for edges

3. Given the following code, what is the shape of the output array edges?

import numpy as np
from scipy import ndimage
img = np.array([[1, 2, 1], [0, 1, 0], [2, 1, 2]])
sobel_x = ndimage.sobel(img, axis=0)
sobel_y = ndimage.sobel(img, axis=1)
edges = np.hypot(sobel_x, sobel_y)

medium

A. (3, 3)

B. (2, 2)

C. (3, 2)

D. (1, 3)

Solution

1. Step 1: Check input image shape

   The input image img is a 3x3 array, so shape is (3, 3).

2. Step 2: Understand Sobel output shape

   Sobel filter preserves the input shape, so sobel_x and sobel_y are also (3, 3).

3. Step 3: Calculate edges shape

   Using np.hypot combines two (3, 3) arrays element-wise, resulting in (3, 3).

4. Final Answer:

   (3, 3) -> Option A

5. Quick Check:

   Input shape = output shape for sobel [OK]

Hint: Sobel keeps input shape; hypot combines same shapes [OK]

Common Mistakes:

• Assuming output shape shrinks after sobel
• Confusing axis with shape dimensions
• Thinking hypot changes array shape

4. The following code attempts to apply the Laplace filter but raises an error. What is the mistake?

from scipy import ndimage
img = [[1, 2, 3], [4, 5, 6], [7, 8]]
laplace_img = ndimage.laplace(img)

medium

A. ndimage.laplace does not exist in SciPy

B. Laplace filter requires axis parameter

C. The image must be grayscale

D. The input image must be a NumPy array, not a list

Solution

1. Step 1: Check input type for ndimage filters

   ndimage functions require NumPy arrays, not Python lists, for image input.

2. Step 2: Identify error cause

   Passing a list causes a TypeError because ndimage.laplace expects an array.

3. Final Answer:

   The input image must be a NumPy array, not a list -> Option D

4. Quick Check:

   ndimage needs np.array input [OK]

Hint: Convert lists to np.array before filtering [OK]

Common Mistakes:

• Passing lists instead of arrays
• Thinking laplace needs axis parameter
• Believing laplace function is missing

5. You want to detect edges in a noisy grayscale image using SciPy. Which approach best combines Sobel and Laplace filters to improve edge detection?

hard

A. Apply Gaussian blur, then Sobel filter on axis=0 only

B. Apply Sobel filters on both axes, combine results, then apply Laplace to sharpen edges

C. Use only Laplace filter because Sobel does not work on noisy images

D. Apply Laplace filter first, then Sobel on axis=2

Solution

1. Step 1: Understand Sobel and Laplace roles

   Sobel detects edges by gradients on horizontal and vertical axes; combining both gives full edge info.

2. Step 2: Use Laplace to enhance edges

   Applying Laplace after Sobel can sharpen edges by detecting second-order changes.

3. Step 3: Evaluate options

   Apply Sobel filters on both axes, combine results, then apply Laplace to sharpen edges correctly combines Sobel on both axes and Laplace for sharpening; others misuse axis or filters.

4. Final Answer:

   Apply Sobel filters on both axes, combine results, then apply Laplace to sharpen edges -> Option B

5. Quick Check:

   Sobel + Laplace = better edge detection [OK]

Hint: Combine Sobel axes then Laplace for sharper edges [OK]

Common Mistakes:

• Applying Sobel on invalid axis
• Skipping combination of horizontal and vertical Sobel
• Using Laplace alone without Sobel for noisy images