The AI Interview - Master AI/ML Interviews

Task: Sobel Edge Detection

Edge detection is a fundamental technique in computer vision used to identify boundaries within images. In this task, you will implement a function sobel_edge_detection(image) that applies the Sobel operator to detect edges in a grayscale image.

Input:

image: A 2D list/array representing a grayscale image with pixel values in range [0, 255]

Output:

Return the edge magnitude image as a 2D list with integer values normalized to [0, 255]
The output dimensions will be (H-2, W-2) due to valid convolution (no padding)
Return -1 for invalid inputs

Edge Cases to Handle:

Input is not a valid 2D array
Image dimensions are smaller than 3x3 (minimum required for Sobel)
Any pixel values are outside the valid range (0-255)
Empty image

Notes:

Use the standard Sobel kernels for gradient computation
Compute gradient magnitude from horizontal and vertical gradients
Normalize the output to the range [0, 255] based on the maximum magnitude value

Examples

Example 1:

Input:

image = [[0, 0, 255], [0, 0, 255], [0, 0, 255]]
print(sobel_edge_detection(image))

Output: [[255]]

Explanation: This 3x3 image has a strong vertical edge (black on left, white on right). The Sobel operator computes: Gx = 1020 (strong horizontal gradient), Gy = 0 (no vertical gradient). The magnitude sqrt(1020^2 + 0^2) = 1020 is normalized to 255 since it's the maximum value.

Starter Code

import numpy as np

def sobel_edge_detection(image):
    """
    Apply Sobel edge detection to a grayscale image.
    
    Args:
        image: 2D list/array representing a grayscale image
               with values in range [0, 255]
    
    Returns:
        Edge magnitude image as 2D list with integer values (0-255),
        or -1 if input is invalid
    """
    # Your code here
    pass