Edge detection (Canny) in Computer Vision - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Predict Output

intermediate

Output of Canny edge detection with different thresholds

Given the following Python code using OpenCV to apply Canny edge detection, what will be the shape of the output image?

Computer Vision

import cv2
import numpy as np

image = np.zeros((100, 100), dtype=np.uint8)
cv2.rectangle(image, (25, 25), (75, 75), 255, -1)
edges = cv2.Canny(image, 50, 150)
print(edges.shape)

A (98, 98)

B (50, 50)

C (102, 102)

D (100, 100)

💡 Hint

Remember that Canny edge detection returns an image of the same size as the input.

The Canny function returns an edge map with the same height and width as the input image. So the output shape is (100, 100).

❓ Model Choice

intermediate

Choosing parameters for Canny edge detection

Which pair of threshold values is most likely to detect edges clearly without too much noise in a typical grayscale image?

A Low threshold = 100, High threshold = 110

B Low threshold = 50, High threshold = 150

C Low threshold = 10, High threshold = 200

D Low threshold = 200, High threshold = 250

💡 Hint

The high threshold should be significantly higher than the low threshold to filter noise.

Typical good values for Canny thresholds are low around 50 and high around 150, balancing edge detection and noise reduction.

🔧 Debug

advanced

Why does this Canny edge detection code raise an error?

Consider this code snippet:

import cv2
image = cv2.imread('image.jpg')
edges = cv2.Canny(image, 100, 200)

It raises an error. What is the cause?

A The image is not converted to grayscale before applying Canny

B The threshold values are invalid (should be between 0 and 1)

C The image path is incorrect

D Canny requires a color image, but a grayscale image was given

💡 Hint

Canny expects a single channel image.

Canny edge detection requires a grayscale image input. If the image is read in color (3 channels), it causes an error.

❓ Metrics

advanced

Evaluating edge detection quality

Which metric is best suited to quantitatively compare the quality of edges detected by Canny against a ground truth edge map?

A Mean Squared Error (MSE)

B Accuracy

C Intersection over Union (IoU)

D Root Mean Squared Error (RMSE)

💡 Hint

Think about how well the detected edges overlap with the true edges.

IoU measures the overlap between predicted edges and ground truth edges, making it suitable for edge detection evaluation.

🧠 Conceptual

expert

Understanding the role of Gaussian smoothing in Canny edge detection

Why does the Canny edge detection algorithm apply Gaussian smoothing before finding edges?

A To reduce noise and avoid detecting false edges

B To increase the contrast of edges

C To sharpen the image details

D To convert the image to grayscale

💡 Hint

Think about what noise does to edge detection.

Gaussian smoothing blurs the image slightly to reduce noise, which helps prevent false edge detection in the next steps.