10 10 20 10 10 20 10 10 20 Gx convolution at center: (-1×10)+(0×10)+(+1×20) + (-2×10)+(0×10)+(+2×20) + (-1×10)+(0×10)+(+1×20) = (-10+0+20) + (-20+0+40) + (-10+0+20) = 10 + 20 + 10 = 40. Gy = 0 (uniform vertically). Magnitude = 40 → strong vertical edge. Q8. Convolution and correlation are identical operations in image processing. Solution: False. In convolution, the kernel is flipped (rotated 180°) before applying; correlation does not flip.
b) Middle value after sorting the 9 neighbors – definition of median filter. Section B: Short Answer Q3. What is histogram equalization? Write its main advantage and one limitation.
Sobel operator approximates gradient using two 3×3 masks:
Extract 3×3 neighborhood around row3,col3 (value=10) – rows 2-4, cols 2-4 (1-indexed): Image Processing Exam Questions And Solutions
| Spatial Domain | Frequency Domain | |----------------|------------------| | Operates directly on pixels | Operates on Fourier transform of image | | Uses masks/kernels (e.g., Sobel, averaging) | Uses filters (low-pass, high-pass) | | Faster for small kernels | Faster for large kernels (using FFT) | | Intuitive for local operations | Better for periodic noise removal | Q5. Given a 5×5 image region (pixel values):
Here’s a useful, structured piece covering for an undergraduate-level Image Processing course. It includes multiple-choice, short answer, and problem-solving formats with explained solutions. Image Processing: Exam Questions & Solutions Section A: Multiple Choice (concepts) Q1. Which operation is not a point operation? a) Log transformation b) Histogram equalization c) Median filtering d) Gamma correction
Final mapping: 0→4, 2→6, 4→8, 6→11, 10→13, 14→15 Q7. Explain the steps to perform edge detection using the Sobel operator. Include masks and a brief example. 10 10 20 10 10 20 10 10
Output pixel = Q6. Perform histogram equalization on a 4-bit image (0-15) with histogram: Gray level: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Frequency: 2 0 1 0 1 0 2 0 0 0 1 0 0 0 1 0 Total pixels = 8
10 12 12 14 16 12 10 12 14 16 12 12 10 14 16 14 14 14 10 18 16 16 16 18 20 Compute the output of a at center position (row 3, col 3) – 1-indexed (value=10). Use zero-padding.
c) Median filtering – it is a spatial operation using a neighborhood, not a point operation. Q2. In a 3×3 median filter applied to a grayscale image, the output pixel value is: a) Mean of the 9 neighbors b) Middle value after sorting the 9 neighbors c) Most frequent value d) Weighted sum of neighbors In convolution, the kernel is flipped (rotated 180°)
10 12 14 12 10 14 14 14 10 Flatten: [10,12,14,12,10,14,14,14,10] Sorted: [10,10,10,12,12,14,14,14,14] Median (5th value) =
| r_k | freq | CDF | CDF_norm = CDF/8 | Equalized = round(15 × CDF_norm) | |-----|------|-----|------------------|----------------------------------| | 0 | 2 | 2 | 0.250 | 4 | | 1 | 0 | 2 | 0.250 | 4 | | 2 | 1 | 3 | 0.375 | 6 | | 3 | 0 | 3 | 0.375 | 6 | | 4 | 1 | 4 | 0.500 | 8 | | 5 | 0 | 4 | 0.500 | 8 | | 6 | 2 | 6 | 0.750 | 11 | | 7 | 0 | 6 | 0.750 | 11 | | 8-14| 0 | 6 | 0.750 | 11 | | 10 | 1 | 7 | 0.875 | 13 | | 14 | 1 | 8 | 1.000 | 15 |