Shaheed Zulfikar Ali Bhutto Institute of Science & Technology University
ROBOTICS & ARTIFICIAL INTELLIGENCE DEPARTMENT
Total Marks: 04
Obtained Marks:
Computer Vision
Assignment # 02
Last Date of Submission: 25th October 2024
Submitted To: Dr. Muhammad Junaid Umer
Student Name: Haisam Abbas
Reg. Number: 23108406
Shaheed Zulfikar Ali Bhutto Institute of Science & Technology University
Instructions: Copied or shown assignments will be marked zero. Late submissions are not
entertained in any case.
Note: Every coding block output must include your name as print function
must submit Jupyter and Pdf files
Take images of your choice and perform the following task
Q1: Take any image and perform smoothing using averaging mean median max
and min filters show the difference(Marks=1.5)
Q2: Take a single image of your choice and perform sharpening using pewit,
and sobel operators and DoG(Marks=1.5)
Q3: After applying the above operations describe the conclusion of each
operation what changes are observed after applying each operation. (Marks=1)
Answer:
Q1: Take any image and perform smoothing using averaging mean median max
and min filters show the difference(Marks=1.5)
# Q1: Image Smoothing Operations
import cv2
import numpy as np
import matplotlib.pyplot as plt
from scipy import ndimage
def print_name():
print("Haisam Abbas: 23108406")
def apply_smoothing(image_path):
print_name()
# Read and convert image
img = cv2.imread(image_path)
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
# Apply filters
# Average filter
avg_kernel = np.ones((5,5))/25
avg_filtered = cv2.filter2D(img, -1, avg_kernel)
# Mean filter
mean_filtered = cv2.blur(img, (5,5))
# Median filter
median_filtered = cv2.medianBlur(img, 5)
# Max filter
max_filtered = ndimage.maximum_filter(img, size=5)
Shaheed Zulfikar Ali Bhutto Institute of Science & Technology University
# Min filter
min_filtered = ndimage.minimum_filter(img, size=5)
# Display results
plt.figure(figsize=(20, 10))
images = [img, avg_filtered, mean_filtered, median_filtered,
max_filtered, min_filtered]
titles = ['Original', 'Average Filter', 'Mean Filter', 'Median
Filter', 'Max Filter', 'Min Filter']
for i in range(len(images)):
plt.subplot(2, 3, i+1)
plt.imshow(images[i])
plt.title(titles[i])
plt.axis('off')
plt.tight_layout()
plt.show()
# Run Q1
print_name()
# Replace 'your_image1.jpg' with your image path
apply_smoothing('/content/boats-in-lagoon-with-noise.jpg')
Output:
Shaheed Zulfikar Ali Bhutto Institute of Science & Technology University
Q2: Take a single image of your choice and perform sharpening using pewit,
and sobel operators and DoG(Marks=1.5)
Answer:
# Q2: Image Sharpening Operations
import cv2
import numpy as np
import matplotlib.pyplot as plt
def print_name():
print("Haisam Abbas: 23108406")
def apply_sharpening(image_path):
print_name()
# Read and convert image
img = cv2.imread(image_path)
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
# Convert to grayscale for edge detection
gray = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
# Prewitt operator
kernelx = np.array([[-1,0,1], [-1,0,1], [-1,0,1]])
kernely = np.array([[-1,-1,-1], [0,0,0], [1,1,1]])
prewitt_x = cv2.filter2D(gray, -1, kernelx)
prewitt_y = cv2.filter2D(gray, -1, kernely)
prewitt = cv2.addWeighted(np.absolute(prewitt_x), 0.5,
np.absolute(prewitt_y), 0.5, 0)
# Sobel operator
sobelx = cv2.Sobel(gray, cv2.CV_64F, 1, 0, ksize=3)
sobely = cv2.Sobel(gray, cv2.CV_64F, 0, 1, ksize=3)
sobel = np.sqrt(sobelx**2 + sobely**2)
sobel = np.uint8(sobel)
# Difference of Gaussians (DoG)
gaussian1 = cv2.GaussianBlur(gray, (3,3), 0)
gaussian2 = cv2.GaussianBlur(gray, (15,15), 0)
dog = gaussian1 - gaussian2
# Display results
plt.figure(figsize=(20, 5))
images = [gray, prewitt, sobel, dog]
titles = ['Original Grayscale', 'Prewitt Operator', 'Sobel Operator',
'Difference of Gaussians']
for i in range(len(images)):
Shaheed Zulfikar Ali Bhutto Institute of Science & Technology University
plt.subplot(1, 4, i+1)
plt.imshow(images[i], cmap='gray')
plt.title(titles[i])
plt.axis('off')
plt.tight_layout()
plt.show()
# Run Q2
print_name()
# Replace 'your_image2.jpg' with your image path
apply_sharpening('/content/Valve_original_(1).png')
Output:
Q3: After applying the above operations describe the conclusion of each
operation what changes are observed after applying each operation. (Marks=1)
Answer:
Smoothing Filters Analysis
1. Average Filter
Effect: Creates uniform blur across the image
Pros:
o Effective for general noise reduction
o Simple implementation
o Uniform smoothing effect
Cons:
o Can significantly blur edges
o Loss of fine details
Best Use Case: When uniform smoothing is desired
Observations:
o Reduces random noise effectively
o Works well for images with subtle texture variations
2. Mean Filter
Effect: Similar to average filter but with better edge preservation
Pros:
Shaheed Zulfikar Ali Bhutto Institute of Science & Technology University
o Computationally efficient
o Good for Gaussian noise reduction
o Better edge preservation than average filter
Cons:
o Still causes some edge blurring
o May not preserve fine details
Best Use Case: Quick noise reduction with moderate edge preservation
Observations:
o Faster processing compared to more complex filters
o Balanced approach between noise reduction and detail preservation
3. Median Filter
Effect: Removes noise while preserving edges better than mean/average
Pros:
o Excellent for salt-and-pepper noise removal
o Better edge preservation
o Good for removing outliers
Cons:
o Can remove fine image details
o Slower than mean filter
Best Use Case: Images with impulse noise (salt-and-pepper)
Observations:
o Very effective for removing random sharp noise
o Maintains sharp edges better than linear filters
4. Max Filter
Effect: Brightens image and enlarges bright regions
Pros:
o Good for removing dark noise
o Enhances bright features
o Useful for peak detection
Cons:
o Can cause bright regions to expand
o May create unwanted bright artifacts
Best Use Case: Removing dark spots or pepper noise
Observations:
o Particularly effective for dark noise removal
o Can help in highlighting bright features
5. Min Filter
Effect: Darkens image and enlarges dark regions
Pros:
o Good for removing bright noise
o Enhances dark features
o Useful for valley detection
Cons:
o Can cause dark regions to expand
o May create unwanted dark artifacts
Best Use Case: Removing bright spots or salt noise
Observations:
o Effective for bright noise removal
Shaheed Zulfikar Ali Bhutto Institute of Science & Technology University
o Can help in highlighting dark features
Sharpening Operators Analysis
1. Prewitt Operator
Effect: Detects edges in horizontal and vertical directions
Pros:
o Simple and effective edge detection
o Computationally efficient
o Good for basic edge enhancement
Cons:
o More sensitive to noise than Sobel
o Less accurate in diagonal directions
Best Use Case: Basic edge detection in low-noise images
Observations:
o Produces thicker edges compared to Sobel
o Works well for clear, well-defined edges
2. Sobel Operator
Effect: Enhanced edge detection with better noise handling
Pros:
o Better noise resistance than Prewitt
o More accurate edge detection
o Good directional sensitivity
Cons:
o May miss some subtle edges
o Slightly more complex computation
Best Use Case: General-purpose edge detection
Observations:
o Produces cleaner edges than Prewitt
o Better performance in noisy conditions
3. Difference of Gaussians (DoG)
Effect: Enhances edges at different scales
Pros:
o Can detect edges at multiple scales
o Good for feature detection
o Less sensitive to noise
Cons:
o May produce thicker edges
o Results depend on chosen Gaussian parameters
Best Use Case: Feature detection and blob detection
Observations:
o Effective for detecting features of various sizes
o Good balance between edge detection and noise suppression
Summary of Changes
After applying these operations, the following key changes were observed:
1. Smoothing Operations:
Shaheed Zulfikar Ali Bhutto Institute of Science & Technology University
oReduced image noise
oDecreased fine detail visibility
oVarious degrees of edge preservation
oDifferent effectiveness for different types of noise
2. Sharpening Operations:
o Enhanced edge visibility
o Increased contrast at edges
o Different sensitivity to noise and directions
o Various scales of feature detection
The End
