Morphological operations (erosion, dilation) in SciPy - Mini Project: Build & Apply

Morphological Operations with SciPy

📖 Scenario: Imagine you have a simple black and white image represented as a grid of 0s and 1s. You want to clean up the image by removing small noise or filling small holes. Morphological operations like erosion and dilation help with this.

🎯 Goal: You will create a small binary image using a NumPy array, set a structuring element, apply erosion and dilation using SciPy, and then display the results.

📋 What You'll Learn

Create a binary image as a 2D NumPy array with exact values

Create a structuring element using SciPy's morphology module

Apply erosion and dilation to the image using the structuring element

Print the original image and the results of erosion and dilation

💡 Why This Matters

🌍 Real World

Morphological operations are used in image processing to clean up images, remove noise, and prepare images for further analysis like object detection.

💼 Career

Understanding and applying morphological operations is important for roles in computer vision, medical imaging, and any job involving image data preprocessing.

1

Create the binary image

Create a 2D NumPy array called image with these exact values: [[0, 0, 1, 1, 0], [0, 1, 1, 1, 0], [1, 1, 1, 1, 1], [0, 1, 1, 1, 0], [0, 0, 1, 1, 0]]

SciPy

# Create the binary image array called image
# Your code here

Hint

Use np.array to create a 2D array with the exact rows and columns shown.

2

Create the structuring element

Import scipy.ndimage as ndi and create a 3x3 structuring element called structure using ndi.generate_binary_structure(2, 1)

SciPy

import numpy as np

image = np.array([[0, 0, 1, 1, 0],
                  [0, 1, 1, 1, 0],
                  [1, 1, 1, 1, 1],
                  [0, 1, 1, 1, 0],
                  [0, 0, 1, 1, 0]])

# Import scipy.ndimage as ndi and create a structuring element called structure
# Your code here

Hint

Use ndi.generate_binary_structure with 2 dimensions and connectivity 1 to create a 3x3 structuring element.

3

Apply erosion and dilation

Use ndi.binary_erosion with image and structure to create eroded. Use ndi.binary_dilation with image and structure to create dilated.

SciPy

import numpy as np
import scipy.ndimage as ndi

image = np.array([[0, 0, 1, 1, 0],
                  [0, 1, 1, 1, 0],
                  [1, 1, 1, 1, 1],
                  [0, 1, 1, 1, 0],
                  [0, 0, 1, 1, 0]])

structure = ndi.generate_binary_structure(2, 1)

# Apply binary erosion and dilation to image using structure
# Your code here

Hint

Use ndi.binary_erosion and ndi.binary_dilation functions with the structure parameter.

4

Print the original and processed images

Print the image, eroded, and dilated arrays each on separate lines using print().

SciPy

import numpy as np
import scipy.ndimage as ndi

image = np.array([[0, 0, 1, 1, 0],
                  [0, 1, 1, 1, 0],
                  [1, 1, 1, 1, 1],
                  [0, 1, 1, 1, 0],
                  [0, 0, 1, 1, 0]])

structure = ndi.generate_binary_structure(2, 1)

eroded = ndi.binary_erosion(image, structure=structure)
dilated = ndi.binary_dilation(image, structure=structure)

# Print image, eroded, and dilated arrays
# Your code here

Hint

Use print() to show each array. The eroded and dilated arrays will show True and False values.

Practice

1. What does the erosion operation do to a binary image in morphological processing?

easy

A. It inverts the colors of the image.

B. It enlarges the white regions by adding pixels to edges.

C. It shrinks the white regions by removing edge pixels.

D. It blurs the image to reduce noise.

Solution

1. Step 1: Understand erosion effect

   Erosion removes pixels on object boundaries, making white regions smaller.

2. Step 2: Compare with other operations

   Dilation adds pixels, inversion changes colors, blurring smooths image; none match erosion.

3. Final Answer:

   It shrinks the white regions by removing edge pixels. -> Option C

4. Quick Check:

   Erosion = Shrink white regions [OK]

Hint: Erosion shrinks shapes by cutting edges [OK]

Common Mistakes:

• Confusing erosion with dilation
• Thinking erosion adds pixels
• Mixing erosion with color inversion

2. Which of the following is the correct way to import the erosion function from scipy's ndimage module?

easy

A. from scipy.ndimage import erosion

B. from scipy.ndimage import binary_erosion

C. import scipy.ndimage.erosion

D. from scipy import erosion

Solution

1. Step 1: Identify correct function name

   The function for erosion on binary images is named binary_erosion in scipy.ndimage.

2. Step 2: Check import syntax

   Correct import is from scipy.ndimage import binary_erosion. Other options are invalid or incorrect names.

3. Final Answer:

   from scipy.ndimage import binary_erosion -> Option B

4. Quick Check:

   Correct import = binary_erosion [OK]

Hint: Use binary_erosion for erosion in scipy.ndimage [OK]

Common Mistakes:

• Using wrong function name 'erosion'
• Incorrect import syntax
• Trying to import from scipy root

3. Given the following code, what will be the output array after applying dilation?

import numpy as np
from scipy.ndimage import binary_dilation

image = np.array([[0, 0, 0, 0, 0],
                  [0, 1, 1, 0, 0],
                  [0, 1, 0, 0, 0],
                  [0, 0, 0, 0, 0]])

result = binary_dilation(image).astype(int)
print(result)

medium

A. [[0 0 0 0 0]
[0 0 1 0 0]
[0 1 0 0 0]
[0 0 0 0 0]]

B. [[0 0 0 0 0]
[0 1 1 0 0]
[0 1 0 0 0]
[0 0 0 0 0]]

C. [[1 1 1 0 0]
[1 1 1 1 0]
[1 1 1 1 0]
[0 1 1 0 0]]

D. [[0 1 1 0 0]
[1 1 1 1 0]
[1 1 1 0 0]
[0 1 0 0 0]]

Solution

1. Step 1: Understand binary_dilation effect

   Dilation adds pixels to the edges of white regions (1s), expanding them by one pixel in all directions.

2. Step 2: Apply dilation to given image

   Original white pixels at (1,1),(1,2),(2,1). After dilation, neighbors become 1, resulting in the array in [[0 1 1 0 0] [1 1 1 1 0] [1 1 1 0 0] [0 1 0 0 0]].

3. Final Answer:

   [[0 1 1 0 0] [1 1 1 1 0] [1 1 1 0 0] [0 1 0 0 0]] -> Option D

4. Quick Check:

   Dilation = expand edges [OK]

Hint: Dilation grows white pixels by one layer [OK]

Common Mistakes:

• Confusing dilation with erosion
• Not converting boolean to int for print
• Misreading array indices

4. The following code is intended to perform erosion on a binary image, but it raises an error. What is the problem?

import numpy as np
from scipy.ndimage import erosion

image = np.array([[1, 1, 0],
                  [1, 0, 0],
                  [0, 0, 1]])

result = erosion(image)
print(result)

medium

A. The function 'erosion' does not exist in scipy.ndimage; use 'binary_erosion' instead.

B. The input image must be float type, not int.

C. The image array shape is invalid for erosion.

D. The print statement syntax is incorrect.

Solution

1. Step 1: Check function availability

   Scipy.ndimage does not have a function named 'erosion'; the correct function is 'binary_erosion'.

2. Step 2: Correct the import and usage

   Replace 'from scipy.ndimage import erosion' with 'from scipy.ndimage import binary_erosion' and call 'binary_erosion(image)'.

3. Final Answer:

   The function 'erosion' does not exist in scipy.ndimage; use 'binary_erosion' instead. -> Option A

4. Quick Check:

   Use binary_erosion, not erosion [OK]

Hint: Use binary_erosion, not erosion function [OK]

Common Mistakes:

• Trying to import non-existent 'erosion'
• Ignoring error messages
• Assuming all morphological functions have simple names

5. You have a noisy binary image with small white dots scattered outside the main object. Which sequence of morphological operations using scipy.ndimage would best remove these small dots but keep the main shape mostly intact?

hard

A. Apply erosion followed by dilation (opening) to remove small objects.

B. Apply dilation followed by erosion (closing) to fill small holes.

C. Apply only dilation to enlarge all white areas.

D. Apply only erosion to shrink all white areas drastically.

Solution

1. Step 1: Understand noise removal goal

   Small white dots are noise; we want to remove them without changing main shape much.

2. Step 2: Choose correct morphological sequence

   Opening (erosion then dilation) removes small objects but keeps main shape. Closing fills holes, not remove dots.

3. Final Answer:

   Apply erosion followed by dilation (opening) to remove small objects. -> Option A

4. Quick Check:

   Opening = erosion + dilation removes noise [OK]

Hint: Use opening (erosion then dilation) to remove small noise [OK]

Common Mistakes:

• Using closing instead of opening for noise removal
• Applying only dilation or erosion alone
• Confusing noise removal with hole filling